WEED SPECIES IDENTIFICATION IN DIFFERENT CROPS USING PRECISION WEED MANAGEMENT: A REVIEW

Abstract

Agriculture plays a vital role in societies and requires research, planning, and execution. It is important to research new trends, scientific methods, and boosters that can give a boost it. The farmer can reduce the amount of workload using some technology which is enhancing the quality of cereal. It is important to identify and growth estimation of weed using deep learning technology in the field of convolution neural networks. This review paper is identifying different types of weeds which are harmful to crop and also identify weed controlling mechanism. It is also useful for researchers to assimilate and clustered the weeds using artificial intelligence techniques and machine learning techniques and to study existing technology of weed detection, which is useful for a researcher can propose new techniques for weed classification and detection. This review paper concise the development of weed detection and classification using the most recent technologies in the field of artificial intelligence and image processing techniques. Concretely, the four procedures, i.e., pre-processing, segmentation, feature extraction, and classification is a part of weed detection and classification were presented in detail. Sooner or later, demanding situations and answers furnished by researchers for weed class and detection inside the subject, together with occlusion and overlap of leaves, varying lighting conditions, and specific growth degrees, have been mentioned.

-

1. Introduction Crop1 production is an important component of agriculture which is responsible for global food management. It requires proper planning and management. Therefore, it is important to invent new trends, scientific methods, and boosters that give a boost to crop production. One of the boosters in this field is soil fertility and its management. The measurement of soil to the right amount of fertilizers or fertilizers can ensure the best results. Information on how to use fertilizer and how to improve the productivity of grain can be readily available to farmers at the right time. This is possible with the use latest technology and technique based on artificial intelligence (AI), machine learning (ML) and deep learning (DL), etc. [Indian govt. Nitti Aayog in its discussion paper on 'National Strategy for Artificial Intelligence 'https://niti.gov.in/national strategy-artificialintelligence on 4th June 2018] Explains that the use of artificial intelligence will increase efficiency at each level of agriculture and also increase the income of farmers along with the productivity of crops. These techniques use ‘image recognition’ as an underlying technology through 'deep learning models'. The same is very crucial in the field of weed detection which will be very fruitful to take necessary steps to improve crop production. There are different varieties of weed that are harmful to crop production and need to be detected in the early stage of growth. The growth of weeds within the crop will affect the basic resources such as water, soil, minerals, fresh air, sunlight, etc which is the basic need of the crop. In recent studies, it has been found that 35% of crops destroyed just due to the growth of different types of weeds in the agriculture field. The main objective of this paper is to study different tools and techniques used by the authors to detect and classify weeds, which are necessary for the assessment of weeds development. Several other computer-oriented techniques such as artificial intelligence, wireless sensor network, some other techniques which improve the quality of agriculture for research also help to researcher. These research papers also briefly describe and maintain the biological method of Weed control strategy such as computer vision technology implemented on the biological method of weed control. For each method described with deficiencies recognized results on insects and plant bacillus, and examples of, and capacity for, integration with biological manipulate. This complete paper is sub-divided into different sections: Section 2 describes related work of the concerned area. The weed classification and control techniques are explained under section 3. Section 4, describes Materials and Methodology. Comparative work is given in section 5 and Section 6 is the conclusive section.

2. Related Work

The recent study and research in the field of agriculture predict the yield of the crop is affected by different factors. The weeds are the foremost factor that could harm crop yield. Therefore, this is the most important task to identify and control the weeds at the early stage of weed growth. This section describes the different types of weeds and their management and control techniques. The weed and control classifications are laid-down below: Yuewei Yang et al. [ 1 ] have suggested the using positive enable technique find the exact location of the object and get solution by an encoderdecoder conventional neural network (CNN) were used for fast weed identification of harmful plan like weeds. Further, Chechliński, Łukasz, and Barbara Siemiątkowska. et al. [ 2 ] suggested clustering methods like weeds segmentation and classification based on deep learning also explain the benefits, the loss has been discussed. Rasti, Pejman et al. [ 3 ] And Inkyu Sa 1,Marija Popovic et al. [ 4 ] discuss some techniques for identifying and detect the weed to increase crop production using image processing. Reduce the weeds using automatic robot technique with semantic segmentation CNN (mobile), feature extraction, and recognition. Aji, Wahyu, and Kamarul Hawari et al. [ 5 ] Briefly stated by this method exclusively classifies the weed using UAV imaginary and transfer learning with FCN technology. Huang, Huasheng, Jizhong Deng et al. [ 6 ] drift the detection of broadleaf weed on various crops. In the weed classification process, an algorithm like multistage scattering transformation was playing an important role, weed detection using convolutional deep learning technique and SVM classifier provide 96.88% accuracy. Zhang, Wenhao, et al. [ 7 ] Has suggested a new architecture of RCNN for classification and detection of weed where weed leaf images were classified by PU learning technique, weed characteristic extricates using positive negative problem technique. The development of remark the broadleaf was typical in the crop, VGGNet model useful for various broadleaf identification like amaranths Viridis, boerhavia diffua, anagallis arvenisis,argemone Mexicana. Jalin Ya and Di Cicco, Maurilio et al. [ 8 ] The weed categorized accuracies were 70.99%, respectively manually weed detection and identification ware time consuming, using Robot technology implemented Tested is modern deep learning-based image segmentation technique differentiate monocot and die cot weed. Huang, W. et al. [ 9 ] Sreelakshmi et al [ 10 ] examine 1119 plants 54 test 682 detection and detection accuracy is 0.37% therefore, some weeds are difficult to distinguish visually. Therefore, the category approach insect the pixelwise object base detection using deep learning VGG 16 FCN technique. Datta et al. [ 11 ] demonstrated a framework to classify weed images. Philipp Lotter et al. [ 12 ] also use pixelwise image segmentation photograph sequences allows our system to robustly estimate a pixelsensible on weed, furnished comparisons to other today's tactics, and display that our device appreciably improves the accuracy of weed segmentation with retraining of the model. Om Tiwari et al. [ 13 ] implements an automated approach for the detection of weeds like transfer learning technique reduce the time for determining the weeds using pertained model implemented on some weeds having better accuracy(90%).Heo Choon Ngo et al. [ 14 ] implement weed detection using color classification, using an automated image classification system is designed using CNN which is distinguish between weeds and crops, also used the robot Lego Mind storm EV3 which is directly connected to the computer will spray weed directly into the area near or at which time the weeds have been detected. Discussed by compare weed detection, deep learning, 10 and 50 meters and implement on machine learning but the image taken with different space. S.Manvel G.Forero et al. [ 15 ]The machine learning technique was obtained 93.23% accuracy as compared to the image processing method. Dyrmann and R. N. Jørgensen et al. [ 16 ] get critical analysis of weed image identification, in this research paper approx 17000 weeds images of the wheat crop, this data has been collected by which ATV-mounted camera, for weed detection implement using fully convolution neural network (FCNN).Nima Teimouri and Mads Dyrmann et al. [ 17 ] Completely focused on weed growth and implement a deep convolution neural network (DCNN) used for weed growth repugnance’s, with the classification of cereal. In this research paper approx 18 weed image species are cover and 9649 images are used for training for the computer system, the computer system can spontaneously, categorized the weed into nine subgroups. That cans performance using of this deep convolution neural network (DCNN) which is estimate 2516 set of images, defluxion of two leaves having 96% accuracy. Andres Milioto, and Philipp Lottes et al.[ 18 ] compartmentalization of the crop and weed in sugar beet plant using deep learning. The stem of the sugar beet image implements a deep convolution neural network (DCNN) scrupulously detecting, the weeds with perception result achieve an average of 96.3%. T. Llorca et al. [ 19 ] identification of weed in tomato plant using transfer learning technique using Google’s inception V3 model, which is used for image classifier provide the accuracy of 88.9%. Oktaviana Rena Indriani et al. [ 20 ] implement GLCM method and Hue, Saturation, Value (HSV) calculations for image process that can calculate and determine the sophistication of tomatoes by using K Nearest Neighbor (KNN), the researcher get complete testing after calculation efficiency rate is 100%, GLCM’ s value is 9. James Perring et al. [ 21 ] Write a survey paper to classify the annual weeds according to 65 scientists from different fields like ecology, taxonomists, etc. Aichen Wanga, c, Wen Zhangb, Xinhua Weia,c, et al. [ 22 ] prepare the review paper which is helpful for researchers, they are implementing using computer visualization with image processing for weed detection, also use the deep conventional machine learning technique. This research paper also helps to prepare for Deerfield Robotics.http://ecoursesonline.iasri.res.in/mod/p age/view.php?id=101845. [ 23 ] This material is weed classification in Weed Management in Horticultural Crops which is completely helpful for the researcher. Lawrence, Wetzel, Arora, et al. [ 24 ] Define aquatic weeds, and classification also explains the ecological compact factor, Among 36 media of 12 aquatic weeds tested for growth of eugenia, worm shows significantly luxuriant growth with Implication of Aquatic Weeds. Manage, S.Abdolrashidi et al. [ 25 ] suggested that two effective sets of abilities had been brought for use for iris recognition: scattering trade-based features and textual content centers. P rosti A. Ahmed S samai ,E bellin,D Russo et al. [ 26 ] recommended that clustering strategies determine the hobby of the lately introduced by.I, B.H .; Zhang, J .; Zheng, W.S. et al. [ 27 ] Rakotomamonjy, A.; Petitjean, C.; Salaün, M.; Thiberville, et al. [ 28 ] Discuss some techniques for identifying and detect the weed for increase the crop production using image processing. Reduce the weeds using automatic robot technique with semantic segmentation CNN (mobile), feature extraction and recognition. Yang, X.; Huang, D.;Wang, Y et al. [ 29 ] briefly stated by this method exclusively classifies the weed using UAV imaginary and transfer learning with FCN technology. Torres-Sánchez, J.;et al. [ 30 ] drift the detection of broad leaf weed on various crops. In the in the weed classification process, an algorithm like multistage scattering transformation was playing an important role, weed detection using conversion machine learning technique and SVM classifier provide 96.88% accuracy. Peña, J.M. ; Torres-Sánchez, J.; Serrano-Pérez, A.; de Castro, et al. [ 31 ] LópezGrandos, F.detection as stricken by quantitative efficacy and sensor resolution. FernandezQuintanilla, C .; Pena, J .; Andujar, D.; Dorado, J .; Ribeiro, A .; Lopez-Grandos, F et al. [ 32 ]the weed categorized accuracies were 70.99%, respectively Manually weed detection and identification ware time consuming, using Robot technology implemented Testbed is a modern deep learning-based image segmentation technique that clearly differentiate monocot and die cot weed. Bakhshipour, A.; Jafari, A. et al. [ 33 ] Aichen Wanga,c, Wen Zhangb et al. [ 34 ] examine 1119 plants 54 test 682 detection and detection accuracy is 0.37%. Therefore, the category approach insect the pixelwise object base detection using deep learning VGG 16 FCN technique. Aichen Wanga,c, Wen Zhangb et al. [ 35 ] demonstrated a framework to classify weed images. Alek‘A r et al. [ 36 ] also use pixel-wise image segmentation photograph sequences allows our system to robustly estimate a pixel-sensible on weed, furnished comparisons to other today's tactics, and display that our device appreciably improves the accuracy of weed segmentation with retraining of the model.Parejo1’ Jin Su Jeong2,Julio HernándezBlanco1 et al. [ 37 ] implement an automated approach for the detection of weeds like transfer learning technique reduce the time for determining the weeds using pertained model implemented on some weeds having better accuracy(90%).Heo choon Ngo et al. [ 38 ] implement weed detection using color classification, using an automated image classification system is designed using CNN which is distinguish between weeds and crops, also used the robot Lego Mind storm EV3 which is directly connected to the computer will spray weed directly into the area near or at which time the weeds have been detected. Discussed by Compare weed detection, deep learning, 10 and 50 meters implement on machine learning but the image taken with different space. M. Ozoemena Ani, Ogbonnaya Onu, Gideon Okoro and Michael Uguru et al. [ 39 ] introduction of biological control method on weeds. Paolo Bàrber et al. [ 40 ] get critical analysis of weed image identification, in this research paper approx 17000 of weeds images of the wheat crop, this data has been collected by which ATV-mounted camera, for weed detection implement using fully convolution neural network (FCNN). Nima Teimouri, and Mads Dyrmann et al. [ 41 ] completely focused on weed growth and implement a deep convolution neural network (DCNN) used for weed growth repugnance’s, with the classification of cereal. In this research paper approx 18 weed image species are cover and 9649 images are used for training for the computer system, the computer system can spontaneously, categorized the weed into nine subgroups. That cans performance using of this deep convolution neural network (DCNN) which is estimate 2516 set of images, defluxion of two leaves having 96% accuracy. Andres Milioto, and Philipp Lottes et al. [ 42 ] Compartmentalization of the crop and weed in sugar beet plant using deep learning. The stem of the sugar beet image implements a deep convolution neural network (DCNN) scrupulously detecting, the weeds with perception result achieve an average of 96.3%. T. Llorca et al. [ 43 ] identification of weed in tomato plant using transfer learning technique using Google’s inception V3 model, which is used for image classifier provides the accuracy of 88.9%. Oktaviana Rena Indiana. et al. [ 44 ] implement GLCM method and Hue, Saturation, Value (HSV) calculations for image process that can calculate and determine the sophistication of tomatoes by using K Nearest Neighbor (KNN), the researcher get complete testing after calculation efficiency rate is 100%, GLCM’ s value is 9. James Perring et al. [ 45 ] Write a survey paper to classify the annual weeds according to 65 scientists from different fields like ecology, taxonomists, etc. Aichen Wanga,c, Wen Zhangb, Xinhua Weia,c, et al. [ 46 ] prepare the review paper which is helpful for researchers, they are implementing using computer visualization with image processing for weed detection, also use the deep conventional machine learning technique. This research paper also helps to prepare for Deerfield Robotics. http://ecoursesonline.iasri.res.in/mod/page/view .php?id=101845. [ 47 ] This material is weed classification in Weed Management in Horticultural Crops which is completely helpful for the researcher. Lawrence, Wetzel, Arora et al. [ 48 ] Define aquatic weeds, and classification also explains the ecological compact factor, Among 36 media of 12 aquatic weeds tested for growth of eugeniae, worm shows significantly luxuriant growth with Implication of Aquatic Weeds. Nima Teimouri, and Mads Dyrmann et al. [ 49 ] completely focused on weed growth and implement a deep convolution neural network (DCNN) used for weed growth repugnance’s, with a classification of cereal. Andres Milioto, and Philipp Lottes et al. [ 50 ] compartmentalization of the crop and weed in sugar beet plant using deep learning. The stem of the sugar beet image implements a deep convolution neural network (DCNN) scrupulously detecting, the weeds with perception result achieves an average of 96.3%.

3. Weeds classifications and Control Methods 2, 50,000 plant species, weeds are approximately 250 species, primary in agricultural and non-agricultural structures. In recent studies, it has been found that the above-described weeds strongly impact on agriculture system which is the result of heavy loss in the agriculture field. Therefore, it is required to identifying, controlling and reducing their impact on the ecosystem

3.1 Weeds classifications

The weed is classified into two categories based on life cycle, habitat, and morphology in figure 1 and based on ecology, soil type, and septicity, and noxious weeds.

I. Based on Life span

a) Annual: Annual weeds life cycle is one year.

This type of weeds like herbs with shallow roots and stems are weak and propagate through seeds. Annual seed after seeding die away and start the production for the next generation of season. There are some most common annual weeds (Table 1). • • •

Monsoon annual: This type of weed's life duration is only four months. E.g Commelina benghalensis, Boerhavia erecta.

Winter annual: These weeds grew during winter sessions and propagate through seeds. Seeding dies away. lambs quarter, Chenopodium album e.g. lambs quarter Summer annual: Kharif corps. Foxtail.

b) Biennial weeds: Biennial weeds life durations for two years. First-year they are simply Vegetative and next year produce the seed and flower. Biennials example: Alternanthera echinata, Daucus carota.

c) Perennials: These types of weeds’ life cycle are more than two years. It has also been categorized into three types.

Simple: Weed born by seed. Eg. Sonchus arvensis Bulbous: Weed propagated from seeds. Eg. Allium sp.

Corm: Plants breed through cream and seeds. Eg. Timothy (Phleum pretense) Monsoon

Annual Commelina Benghalensis Boerhavia Erecta

Corm Perennials: Timothy (Phleum Pratense) Japanese knotweed

Leafy spurge Simple Perennials Sonchus Arvensis

Perennials weeds Bulbous Perennials Allium Sp.

Echinata,

Bermuda gras

Hedge bindweed Daucus Carota

Wild onion

Yarrow Immersed weeds: This type of weeds completely grew up under the water and root in the mud. E.g. Nelumbium speciosum, Jussieua repens.

II. Habitat weeds

Terrestrial weed: That type of weed grew on land soil, called terrestrial plants. The examples of some terrestrial plants are as follows: e.g Air potato Aquatic weeds: Aquatic weed plants grew under the water and complete at least one or more years in a biological clock called aquatic weeds. It is also divided into four subcategories like submerged, emerged, marginal and floating weeds (Table 2).

Submersed weeds: In general, weeds have grown under the water and stems and leave underneath the water facial. Example: Lemma, polyrrhiza e.g. Ceratophyllum demersum.Ceratophyllum Australe Griseb, Ceratophyllum demersum L. (rigid hornwort or common hornwort) - cosmopolitan These cause livestock to the animals that are accumulated along with barley and maintain to farm animals or even as grazing the cattle devours this toxic plant life e.g. fastuosa (L.) Danert , Stramonium fastuosum (L.) are noxious to living things. The berries of Withania somnifera and seeds of Abrus precatorius are poisonous.

Parasitic weeds Parasitic weeds are probably a mixture; the weeds depend entirely on the host plant, the parasites that attack.

Some parasites as given below:

Total root parasite – This type of plant depends on another plant and gets nutrition from them. Dendrophthoe,Orobanche ,Viscum,Santalum Aeginetia, lathrea, cistanche etc.

Partial root parasite – e.g Sandalwood tree, Witch weed, Rhinanthus.

Total stem parasite – e.g Dodder (Cucuta) Cucuta rootless yellow color.

Partial stem parasite – e.g Viscum and Loranthus.

Aquatic weeds: Aquatic weed plants develop in water and have a life-cycle of at least of years and are classified into four types such as submerged, emerged, marginal and floating weeds.

VII. Noxious Weeds:

A poisonous or noxious weed plant discretionary characterized as being particularly unwanted, inconvenient, and hard to control. The status of a plant as a poisonous weed will shift with the lawful translation of a nation or a state, just as with the advancement of new weed control advances. The toxic weeds have a huge ability to imitate and scatter, and they embrace precarious approaches to resist the man's endeavors to dispose of them. The poisonous weeds are some of the time additionally alluded to as exceptional weeds and offensive weeds. Noxious weeds in India Cyperus rotundus, Cynodon dactylon, Parthenium hysterophorus, Eichhornia crassipes, Solanum elaegnifolium, and Orobanche spp.

VIII. Grassland Weeds:

As the name shows, weeds having a place with this class attack prairie, rangelands, and changeless fields, which offer an unexpected biological condition in comparison to the harvest lands. The significant contrast between the two circumstances, from the viewpoint of perspective on weeds, is that while croplands are much of the time worked and upset, the meadows stay undisturbed for an extensive stretch. The meadow weed species, be that as it may, must withstand visit munching, and cutting, just as stomping on by the creatures. Some grassland weeds are equipped with mechanisms to keep the animals away, like bitter leaves, poisonous foliage, prickly shoots, and hard stems.

3.2 Weeds

Methods:

Prevention

Farmers increase crop production if they remove the weed farm crop, for this, they use a weed removal technique, which is based on the ecological theme. From crop management to complete weed management. For example, weed management with low nutrient management. External input system.

These weed identification and control techniques are classified into different categories as given below in Figure 2.

I. Preventive Methods

a) Crop Rotation: The crop rotation is a traditional technique implemented by farmers for increase the productivity and prevents the weed from the crop, simply means different crop grew in the same field known as preventive weed control. There are some weed control method repeated yearly, given in table 3 II. Biological Control Method The biological influences approach uses naturally occurring enemies of the invasive plant to help minimize its effects. Its objective is to Resume the weeds through its herbal opponents and attain permanent weed management hose herbal images of weeds from field, the weeds image acquisition this parts the weed plant image continuously capture images by the camera with high frame rate and resolution and data pre-processing with output images and feature extraction, detection for use the color analysis use is Hue, Saturation, Value (HSV) computation can be used for categorized to determine the mellowness level of weeds. Even though inside the long run, organic management can be cost-effective and diminish the prerequisite Control practices, now not all weeds are suitable for organic manipulation.

III. Cultural Control Method:

The cultural method is commonly related to farming systems, even though a few factors apply to landscape and bush care practices. This may include the usage of plant species that overwhelm other plant means poisons.

IV. Physical Control

Physically control is the elimination of the weeds using physical or mechanical machines. The approach used often depends on the place of weeds to be controlled by a mechanical method, burning or with the aid of hand, etc.

V. Chemical control

The farmer can use the chemical to remove the weeds but it will also affect to soil and crop, although the usage of chemicals isn't continually essential, herbicides can be a vital and powerful aspect of herbicides. 4. Materials and Methodology This section laid-down the methodology and techniques of weed recognition and grouping as

Image segmentation classification and

This is used to extract and classify based on image attributes. In the first process, the images are captured by a digital camera stored. PNG, JPG, JPEG, etc. The image acquisition involved three steps for pre-processing [ 20 ]. The first steps are involved in the RGB image to grayscale images and second, Steps include the resize image and finally filter the image [ 21 ]. Segmentation provides the solution to the image problem, each leaf has a separate feature that significant information is completely helpful to the developer which is recognized and classified. The GLCM is the methods used for texture analysis this degree is accomplished to give the characteristics or reputation of each photo on the way to be used for training and testing [ 22 ].GLCM is a group of patterns that can be used to discover or classify various capabilities of your application, with the help of a recognition system (for example, an ANN). First converts the RGB picture to HSV. Later, it is important to scale the HSV matrix to values between 0 and 64. It occurs that the coincidence matrix is computationally viable prefer Next, that can compute the co-prevalence matrix for the H, S, and V matrices. Thus, you'll have three co-occurrence matrices, and it could be set parameters (entropy, variance) for each of these matrices. It’s far essential to set up correlations between the parameters, to determine which ones are applicable [ 23 ]. The GLCM houses of a photo are expressed as a matrix with the identical wide variety of rows and columns gray price in the photograph. The elements of this matrix rely on the frequency of two detailed pixels. Both Pixel pairs can vary relying on their community. These matrix elements consist of 2d-order statistical [ 24 ].The implementation of segmented images that can be transformed onto a gray level run length matrix. [ 25 ]. Flow chart of GLCM is given below in figure 4. Start

Gray afterRGB scale

Gray scale Matrix pixel GLCM calculation End

4.2. Convolution Layers

This review paper implements Transfer learning the use with Convolution Neural Network (CNN) for weed detection [ 13 ]. Convolutional neural networks use to perform some operations on images and extract some useful information for rained the model. The neural networks are a collection of layers of neurons that are interconnected and the outcome represents the estimates. A Convolution neural network is different and contains three dimensions such as width, height, and depth [ 14 ].

a) Network Architecture: used to become aware of the phototype from an actual photograph. In this is research paper CNN apply to pick out the weed with the category. there are many exclusive forms of photo category technique consist of a huge wide variety of facts set like photograph net[ 11 ], The pre-trained networks which include the VGGNet [ 8 ], AlexNet [ 9 ], GoogLeNet [ 10 ], ResNet [ 11 ] [ 17 ]. Some other image data set provided by a digital camera set on (MAV) Micro Arial Vehicle [ 1 ], Unmanned Arial Vehicle (UAV) [ 2 ], TAV mounted camera, which generates the digital image of weeds, transfer learning use with CNN for weed image classifier, some example listed as in table 4.1 1

Data acquisition Images Pre-Processing and techniques Classifiers

Classified Weed Hue Saturation Value (HSV) color has 3 elements, known as Hue (H), Saturation (S), and cost (V) [ 19 ]. The HSV consists of 3 elements, wherein Hue represents coloration, dyeing for saturation brightness and fee degrees, dominance and brightness degrees. [ 20 ]. The second segment of weed detection recalls the classifiers, the weed diction is two-step trouble i.e. weed and grain plant. [ 27 ].Another classification method also includes K-nearest neighbor (KNN), Complex Tree, and Logistic Regression [ 36 ]. There are transfer methods for the switch of knowledge among Explanation Weed plant image acquisition by the camera with a high frame rate and resolution Feature extraction, detection for use of the GLCM, and HSV method provide high evaluation correlation and homogeneity of pictures.

Conventional neural network (CNN), Pooling, flattering, transfer learning to use with CNN for weed image classifier It can classify die cot and monocot and broadleaf crop weed, for example, Cyprus, Amaranth human beginners. Transfer learning is ordinarily utilized in computer imaginative and prescient and herbal language processing obligations like sentiment evaluation due to the huge amount of computational power required [ 37 ].

4.3. Transfer Learning Approach

a) Training to reuse

b) Using a Pre-Trained Model

c) Feature Extraction

Transfer learning isn't sincerely a device gaining knowledge of approach but may be visible as a "layout technique" within the subject, as an instance, active getting to know. It’s also now not a one-of-a-kind part or takes a look at-place of device mastering [ 38 ]

Computerized identification and selective spraying of weeds (such as dock) in the grass can provide very considerable long-time period ecological and cost advantages. Although the device imaginative and prescient (with the interface to appropriate automation) affords a powerful means of achieving this, the associated demanding situations are bold, because of the complexity of the pics.

Any weed Localization And classification of weed using a scattering transformation Detection of weeds inside the direction of extreme density existence plant life from the pinnacle view in-depth snapshots.

An annotated artificial statistic-set was positioned underneath the size of an employer and a simulator is proposed for a reproducible technique.

Detection of extensive grass weeds in turf grass the use of VGGNet became an exquisite model for detecting various broad floor weeds that develop in Bermuda grass and the detection of cutleaf knight-primrose (Panthera laciniate Hill) in Bahia grass. DatetechNet changed to a high-quality version to be carried out. The mastering fee coverage exponentially decays.

Use image data taken from the unmanned aerial car (UAV) for mapping the weed and crop with a deep neural network (DNN).

5. Comparative Analysis

Selective weeding measures are a critical step in self-sufficient crop control associated with crop health and yield. but, an important venture is to discover dependable and correct weeds to limit harm to the encompassing plants. In this paper, we gift a method for dense semantic weed sorts with multispectral pix amisped through a micro aerial automobile (MAV). In tomatoes leaf has one-of-a-kind maturity level; consequently, it's far necessary to apprehend the proper Sample to decide the extent of maturity. Texture evaluation may be processed with the use of the grey level Coincidence Matrix (GLCM) technique.

This analysis more increased weed detection employing a ground-primarily based mostly system ingenious and anchoring and image processing techniques Crop weed image mapping using water shade method with different species.

Weed Maps, using thresh holding method for weed control in early climates, inflicting star reflections and troubles.

Crop and weed classification in the soybean plant In most cases, weed management within the traditional method depends on manual labor. This method takes time, contributes to high costs, and vital yield losses. The standard application of chemical weed management, however, goes against the hassle of sustainability. To handle this use of laptop computer imaginers and anchors, preciseness agricultural researchers have used remote sensing Weed Maps, Cascaded CNN and SegNet, MAV (Micro Arial Vehicle) SegNet, MAV (Macri Arial Vehicle Under the Curve (AUC) Achieve _ 0.8 F1-score and get 0.78 regions underneath the curve (AUC) class metrics Cascaded CNN and SegNet, MAV (Micro Arial Vehicle) SegNet, MAV (Macro Arial Vehicle Under the Curve (AUC) Achieve _ 0.8 F1-score and get 0.78 regions underneath the curve (AUC) class metrics.

K-NN, and HSV

GLCM,

GLCM and HSV color space technique Weed discrimination using CNN CNN and SegNet, MAV (Micro Arial Vehicle CNN and SegNet, UAV (Unmanned Arial Vehicle CNN EfficientNet Deep convolutional neural networks (DCNN) Weed identification and discrimination from the crop plant Keras(Tensor Flow FW ) VGG-16, DenseNetSVM: 98% VGG-16, DenseNet Use color space techniques like GLCM and HSV color space technique accuracy rate 100%.

Using DCNN got 94% accuracy. 91% 90.08% (DenseNet) Crop and weed image mapping using machine learning UAV remote sensing Using Deep CNN find the Weed growth competition from the crop.

Growth of weeds in young coffee plants using CNN.

Multiclass weed Species data set Dataset for Deep learning: Deep Weeds Weed classification: using Image Net Weed Management in the transition to Conservation Agriculture: a Review Applications of Computer Vision in Plant Pathology: however, this has become mostly useless for weed control in early climates, inflicting star reflections and troubles. Satellite imagination includes cloud cowls.

Weed control pest on rabi crop session weeds using image segmentation method for dense semantic weed sorts with multispectral pix through a micro aerial automobile (MAV).

This looks at situ images involving 18 weed species grown within a Time,8000 leaves of these drawings were used for the trained of the weed statistics is taken from the rabi crop.

Competition of weed in flower plants using ResNet-50 in CNN architectures study technique of interference between flowers. Soil types, photograph judgments, and lighting fixtures situations. The common ordinary performance of this method met the maximum accuracy of 90.79% Survey paper measure completely different views equivalent to implications for regulation of weeds, terrestrial weeds, and annual weed The weed management practices used by farmers in conservation agriculture and the modifications initiated thru its adoption.

Real-time selection support gadgets can beautify crop or plant boom, consequently, increase their productivity, best and financial value. It also permits the North American nation to serve the character by watching plant growth in equalization the surroundings. pc inventive and presenter technology has valid to play a vital place among the degree of programs equivalent to medicine, defense, agriculture, remote sensing, enterprise analysis, etc.

R-CNN machine learning Deep learning CNN (DL‐CNN) models and Faster R-CNN Machine learning, Deep learning CNN (DL‐CNN) models and Faster R-CNN Machine learning, Statically analysis of data taken from the field Deep Learning CNN (DL-CNN) model and quicker R-CNN machine learning, Survey paper on Weed classifications Deep Convolutional Neural Networks (DCNN) [ 10 ] [ 12 ] [ 18 ] [ 20 ]

6. Conclusions

This paper explains different categories of weeds and control methods used in crops. This paper also explores different techniques that are used for weed management such as artificial intelligence, machine learning, and deep learning with their pros and cons. This paper explains different steps to detect and analyze weed-based images. The steps are pre-processing, Classification, identification of crop weed and cereal categorization using image processing, artificial intelligence, and deep Learning processes techniques. In this paper, the evaluation and assessment of various methodologies are mentioned. The emerging approach CNN with transfer learning ideas can be included right into a speaking device that could similarly help farmers in identifying crop weeds of plants. The precise category model allows in predicting the species of pest. Deep learningbased destiny work can gain to farmers. In this review, the paper assists the researcher in further weed identification and detection.

References

[1]. Yang, Yuewei, Kevin J. Liang , and Lawrence Carin . "Object detection as a positiveunlabeled problem." arXiv preprint arXiv: 2002 . 04672 ( 2020 ).

[2]. Chechliński, Łukasz, Barbara Siemiątkowska , and Michał Majewski . "A System for Weeds and Crops Identification-Reaching over 10 FPS on Raspberry Pi with the Usage of MobileNets, DenseNet , and Custom Modifications . " Sensors 19 , no. 17 ( 2019 ): 3787 .

[3]. Rasti, Pejman, Ali Ahmad, Salma Samiei, Etienne Belin, and David Rousseau . "Supervised image classification by scattering transform with application to weed detection in culture crops of high density." Remote Sensing 11 , no. 3 ( 2019 ): 249 .

[4]. Inkyu Sa 1 , Marija Popovi ´c 1, Raghav Khanna 1 ,' WeedMap: A Large-Scale Semantic Weed Mapping Framework Using Aerial Multispectral Imaging and Deep Neural Network for Precision Farming' 27 July 2018; Accepted: 26 August 2018 ; Published: 7 September 2018pp . 1 - 6 ( 2018 ).

[5]. Aji, Wahyu, and Kamarul Hawari . "A Study of Deep Learning Method Opportunity on Palm Oil FFB (Fresh Fruit Bunch) Grading Methods." In 2019 Ahmad Dahlan International Conference Series on Engineering and Science (ADICS-ES 2019) , pp. 22 - 25 . Atlantis Press, 2019 .

[6]. Huang, Huasheng, Jizhong Deng, Yubin Lan, Aqing Yang, Xiaoling Deng , and Lei Zhang . "A fully convolutional network for weed mapping of unmanned aerial vehicle (UAV) imagery." PloS one 13 , no. 4 ( 2018 ): e0196302 .

[7]. Zhang, Wenhao, Mark F. Hansen , Timothy N. Volonakis , Melvyn Smith , Lyndon Smith , Jim Wilson, Graham Ralston, Laurence Broadbent, and Glynn Wright . "Broad-leaf weed detection in pasture." In 2018 IEEE 3rd International Conference on Image, Vision and Computing (ICIVC) , pp. 101 - 105 . IEEE, 2018 .

[8]. Di Cicco, Maurilio, Ciro Potena, Giorgio Grisetti, and Alberto Pretto . "Automatic model-based dataset generation for fast and accurate crop and weeds detection." In 2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pp. 5188 - 5195 . IEEE, 2017 .

[9] . Huang , W. , et al.' New optimized spectral indices for identifying and monitoring winter wheat diseases . ' Appl. Earth Obs. Remote Sens . 7 ( 6 ), 2516 - 2524 ( 2014 )

[10]. Sreelakshmi, M. , Padmanayana: ' Early Detection and Classification of Pests Using Image Processing ' pp. 239 - 242 ( 2015 ).

[11]. Dutta, A. V. I. S. H. E. K. , JOSEPH M. Gitahi , P. R. A. S. H. A. N. T. Ghimire , R. O. B. I. N. Mink , G. Peteinatos, J. Engels , M. Hahn , and R. Gerhards . "Weed detection in close-range imagery of agricultural fields using neural networks . " Publ. DGPF 27 ( 2018 ): 633 - 645

[12]. Lottes, Philipp, Jens Behley, Andres Milioto, and Cyrill Stachniss . "Fully convolutional networks with sequential information for robust crop and weed detection in precision farming . " IEEE Robotics and Automation Letters 3 , no. 4 ( 2018 ): 2870 - 2877 .

[13]. Tiwari, O. , Goyal , V. , Kumar , P. , and Vij , S. , 2019 , April. An experimental set up for utilizing the convolutional neural network in automated weed detection . In 2019 4th International Conference on Internet of Things: Smart Innovation and Usages (IoT-SIU) (pp. 1 - 6 ). IEEE.

[14]. Dasgupta, Ishita, Jayit Saha, Pattabiraman Venkatasubbu, and Parvathi Ramasubramanian . "AI Crop Predictor and Weed Detector Using Wireless Technologies: A Smart Application for Farmers." Arabian Journal for Science and Engineering ( 2020 ): 1 - 13 ..

[15]. Manuel G. Forero1 (&),Sergio HerreraRivera1, 'Color Classification Methods for Perennial Weed Detection in Cereal Crops Systems' (ICACCS), pp. 1 - 4 ( 2017 ).

[16]. Dyrmann, Mads, Rasmus Nyholm Jørgensen, and Henrik Skov Midtiby. "RoboWeedSupportDetection of weed locations in leaf occluded cereal crops using a fully convolutional neural network." Adv. Anim. Biosci 8 , no. 2 ( 2017 ): 842 - 847 ..

[17]. Teimouri, Nima, Mads Dyrmann, Per Rydahl Nielsen, Solvejg Kopp Mathiassen, Gayle J. Somerville , and Rasmus Nyholm Jørgensen. "Weed growth stage estimator using deep convolutional neural networks . " Sensors 18 , no. 5 ( 2018 ): 1580 .

[18]. Milioto, Andres, Philipp Lottes , and Cyrill Stachniss . "Real-time blob-wise sugar beets vs weeds classification for monitoring fields using convolutional neural networks." ISPRS Annals of the Photogrammetry , Remote Sensing and Spatial Information Sciences 4 ( 2017 ): 41 .

[19]. Sladojevic, S. , Arsenovic , M. , Anderla , A. : 'Deep neural networks based recognition of plant diseases by leaf image classification'Comput . Intell. Neurosci. 2016 ( 2016 ).

[20]. Indriani, Oktaviana Rena, Edi Jaya Kusuma, Christy Atika Sari, and Eko Hari Rachmawanto. "Tomatoes classification using K-NN based on GLCM and HSV color space." In 2017 international conference on innovative and creative information technology (ICITech) , pp. 1 - 6 . IEEE, 2017 .

[21]. Pheloung, P. C. , P. A. Williams , and S. R. Halloy . "A weed risk assessment model for use as a biosecurity tool evaluating plant introductions . " Journal of environmental management 57 , no. 4 ( 1999 ): 239 - 251 .

[22]. Janowczyk, Andrew, and Anant Madabhushi . "Deep learning for digital pathology image analysis: A comprehensive tutorial with selected use cases . " Journal of pathology informatics 7 ( 2016 ).

[23]. Weiss, Karl, Taghi M. Khoshgoftaar , and DingDing Wang . "A survey of transfer learning . " Journal of Big data 3 , no. 1 ( 2016 ): 9 .

[24]. Bruna, Joan, and Stéphane Mallat . "Invariant scattering convolution networks . " IEEE transactions on pattern analysis and machine intelligence 35 , no. 8 ( 2013 ): 1872 - 1886 .

[25]. Minaee, Shervin, AmirAli Abdolrashidi, and Yao Wang . "Iris recognition using scattering transform and textural features." In 2015 IEEE signal processing and signal processing education workshop (SP/SPE), pp. 37 - 42 . IEEE, 2015 .

[26]. Rasti, Pejman, Ali Ahmad, Salma Samiei, Etienne Belin, and David Rousseau . "Supervised image classification by scattering transform with application to weed detection in culture crops of high density." Remote Sensing 11 , no. 3 ( 2019 ): 249 .

[27]. Li, Bran Hongwei , Jianguo Zhang, and WeiShi Zheng. "HEp-2 cells staining patterns classification via wavelet scattering network and random forest." In 2015 3rd IAPR Asian Conference on Pattern Recognition (ACPR) , pp. 406 - 410 . IEEE, 2015 .

[28]. Rakotomamonjy, Alain, Caroline Petitjean, Mathieu Salaün, and Luc Thiberville . "Scattering features for lung cancer detection in fibered confocal fluorescence microscopy images . " Artificial intelligence in medicine 61, no. 2 ( 2014 ): 105 - 118 .

[29]. Yang, Xudong, Di Huang, Yunhong Wang , and Liming Chen . "Automatic 3d facial expression recognition using geometric scattering representation." In 2015 11th IEEE International Conference and Workshops on Automatic Face and Gesture Recognition (FG) , vol. 1 , pp. 1 - 6 . IEEE, 2015 .

[30]. Torres-Sánchez, Jorge, Francisca LópezGranados, Ana Isabel De Castro, and José Manuel Peña-Barragán. "Configuration and specifications of an unmanned aerial vehicle (UAV) for early site specific weed management." PloS one 8 , no. 3 ( 2013 ): e58210 .

[31]. Peña, J. M. " Torres-Sa nchez J , SerranoPėrez A , de Castro AI , Lopez-Granados F . Quantifying efficacy and limits of unmanned aerial vehicle (UAV) technology for weed seedling detection as affected by sensor resolution." Transactions of the ASAE 15 , no. 3 ( 2015 ): 5609 - 5626 .

[32]. Fernández‐Quintanilla, C. , J. M. Peña , D. Andújar , J. Dorado , A. Ribeiro , and F. López‐Granados . "Is the current state of the art of weed monitoring suitable for site‐specific weed management in arable crops?" Weed research 58 , no. 4 ( 2018 ): 259 - 272 .

[33]. Bakhshipour, Adel, and Abdolabbas Jafari . "Evaluation of support vector machine and artificial neural networks in weed detection using shape features." Computers and Electronics in Agriculture 145 ( 2018 ): 153 - 160 .

[34]. https://www.intechopen. com/books/biolog ical-approaches-for-controllingweeds/overview-of-biological-methods-ofweed-control

[35]. Wang, Aichen, Wen Zhang, and Xinhua Wei . "A review on weed detection using groundbased machine vision and image processing techniques." Computers and electronics in agriculture 158 ( 2019 ): 226 - 240 .

[36]. 'Weed Classifications' https://www.intechopen.com/books/biologic al-approaches -for-controllingweeds/overview-of-biological-methods-ofweed-control.

[37]. 'Weed Classifications' https://link.springer.com/chapter/10.1007%2 F978 -94-011-4014- 0 _ 7

[38]. Alek'A biological control of weed management' Proc.VII.Int.Symp.Biol.Contr.We eds.6-11 March 1988 ,Rome,Italy,Delfosse,E.S(Ed.).1st, Sper ,Pat ol,Veg,(MAF),PP, 101 - 6 ( 1989 ).

[39]. Akalin, Aysu, Kemal Yildirim, Christopher Wilson, and Onder Kilicoglu . "Architecture and engineering students' evaluations of house façades: Preference, complexity, and impressiveness . " Journal of environmental psychology 29 , no. 1 ( 2009 ): 124 - 132 .

[40]. Green, S. " A review of the potential for the use of bioherbicides to control forest weeds in the UK . " Forestry 76 , no. 3 ( 2003 ): 285 - 298 .

[41]. Ascard, J. , P. E. Hatcher , B. Melander , M. K. Upadhyaya , and R. E. Blackshaw . "10 Thermal weed control." Non-chemical weed management: principles, concepts, and technology ( 2007 ): 155 - 175 .

[42]. Melander, Bo, Ilse A. Rasmussen , and Paolo Bàrberi . "Integrating physical and cultural methods of weed control-examples from European research." Weed Science 53 , no. 3 ( 2005 ): 369 - 381 .

[43]. Melander, Bo, Ilse A. Rasmussen , and Paolo Bàrberi . "Integrating physical and cultural methods of weed control-examples from European research." Weed Science 53 , no. 3 ( 2005 ): 369 - 381 .

[44]. Syrett, P. , D. T. Briese, and J. H. Hoffmann . "Success in biological control of

[45]. terrestrial weeds by arthropods." In Biological control: measures of success , pp. 189 - 230 . Springer, Dordrecht, 2000 .

[46]. Barth, Ruud, Joris IJsselmuiden, Jochen Hemming, and Eldert J. Van Henten . "Data synthesis methods for semantic segmentation in agriculture: A Capsicum annuum dataset." Computers and electronics in agriculture 144 ( 2018 ): 284 - 296 .

[47]. https://niti.gov.in/writereaddata/files/doc ument_publication/NationalStrategy-for-AIDiscussion-Paper.pdf.

[48]. De Castro, Ana I. , Jorge Torres-Sánchez, Jose M. Peña, Francisco M. Jiménez-Brenes , Ovidiu Csillik , and Francisca López-Granados. "An automatic random forest-OBIA algorithm for early weed mapping between and within crop rows using UAV imagery." Remote Sensing 10 , no. 2 ( 2018 ): 285 .

[49]. Sa, Inkyu, Marija Popović, Raghav Khanna, Zetao Chen, Philipp Lottes, Frank Liebisch, Juan Nieto, Cyrill Stachniss, Achim Walter , and Roland Siegwart . "Weed map: a largescale semantic weed mapping framework using aerial multispectral imaging and deep neural network for precision farming . " Remote Sensing 10 , no. 9 ( 2018 ): 1423 .

[50]. Zhang, Chunhua, and John M. Kovacs. " The application of small unmanned aerial systems for precision agriculture: a review." Precision agriculture 13 , no. 6 ( 2012 ): 693 - 712 .

[51]. ÓPEZ‐GRANADOS, Francisca. "Weed detection for site‐specific weed management: mapping and real‐time approaches . " Weed Research 51 , no. 1 ( 2011 ): 1 - 11 . https://niti.gov.in/nat