The article gives a description of a package solution for monitoring and control of forestry enterprise activity. The solution consists of manufacturing execution system (MES) and mobile application for roundwood volume control in situ. The system enables on-line workflow monitoring at the low landing as well as roundwood acceptance and shipment on line monitoring. This solution provides full automation of gathering data about raw material logistic at each processing stage till delivery.
Sales department Directory of clients and contracts; production plan Forest workers
Management Insights about current condition and activity of the roundwood yard Dispatch reports
Control and accounting of the forest enterprise activity
Step-by-step data about roundwood yard activity
Roundwood yard
specialists Production directory
Side rods
Planning department of forest exploitation
Chew directory and stuff composition
Registration of the income and dispatch
of the timber
Forest use reports
Fig. 1 –Model of the automation objects Jumbled timber income
These operations are stored in form of specifications. The specifications are uploaded to the system’s server in the .json text format. The transmitted package also includes inner image which is used for the measurement and further visual validation of the volume and quantity of the processed roundwood. The example of the package is listed below. { "datetime": "2017-04-19T07:55:45", "device_id": "54b1ea3ca62185bc", "dictionaries": { "contract_sale": "e17ec2f1-833f-443b-8116-aaa91e97118c", "depot_master": "841c3aa8-5fdd-460c-9fa6-a958a41049f0", "driver": "c93a8e70-dd42-4958-8794-2439fc34fae2", "forest_declaration_area": "6e193c4c-9f60-48b1-a8fb-0d9a23f9b072", "load_operator": "aa313f4c-4327-4a44-9d02-e8fcd69a75fd", "outcome_type_enum": "Log truck", "railway_carriage_text": "", "railway_route_text": "", "team_sortiment": "2f288591-9fb9-4a57-a109-61ca85ded806", "team_wood": "", "unload_operator": "0475a24e-99fc-4dd7-8442-bc1df3b5ed05", "wood_outcome_deriction_enum": "", "wood_sortiment": "50c53bb9-3144-4a91-9590-ee14c0966071", "wood_type": "", "wood_type_enum": "", "woodcutter": "" } }, "images": ["image array"], "operation_type": 0, "reporter_name": "Last name, First name", "specific_id": "1492588545754", "specific_name": "Specification", "sum_logs": 0, "sum_volume": 0.0, "timbers": []
The data exchange between “FoRest” software and the main system is realized through assigned in the system’s settings catalogue titled “FoRest downloads” – that is network resource for downloading files from the software. This folder must include folder “processed” for storing files which are processed and taken into account by the system. 2. Round timber automatic measurement “FoRest” is designed to conduct operations with maximum possible automatic performance, however, it provides tools for manual editing of the processing result. The software is designed for mobile devices under Android OS ver. 4.3 or above. The sequence of operations for obtaining the log pile measurement results is following: 1) Uploading one or two images into the program. Number of images is selected according to the available viewpoints of the log pile ends and assigned in the global settings of the program 2) Calibration of each image. Calibration consists in determination of the inner and outer parameters of the camera. It can be implemented in automatic mode with particular standard object detection or manually by the user. 3) Automatic detection of abuts. This stage is performed independently from the user 4) Manual editing. It is implemented in cases of algorithm detection error, for example for the overlapped abuts. 5) Result generation and analysis. Program output is presented in the form of a detailed report which can be edited afterwards or exported in doc, xls, pdf format.
One of the most common deterministic methods is the Hough transform (HT) method. On the basis
of this method, the algorithm of probabilistic pair voting (PPV) for fast and reliable detection of
circular objects is proposed [
Other methods for recognizing ellipses include the static RANSAC method described in [
The method proposed in [
In [
The best result during the tests were obtained by ELSD algorithm [
The ELSD algorithm is a three-step process: 1) at the first stage of the heuristic method - candidates for the desired figures; 2) then each candidate goes through a validation phase (validation of candidates); 3) Finally, to select the best geometric interpretation, a model selection stage is required.
The first step is based on the greedy (heuristic) approach, which was proposed in the algorithm for
recognizing LSD segments [
Gray-level image X grad = compute_gradient(X)
foreach x[i] in X R = region_grow(x[i],grad) C = curve_grow(R,grad) line = fit_rectangle(R) circle = fit_circular_ring(C) ellipse = fit_elliptical_ring(C) No (nfa_l,nfa_c,nfa_e) = NFA(line,circle,ellipse)
NFA_min = min(nfa_l,nfa_c,nfa_e)
NFA_min <= 1
Yes L = NFA_min
List L
Fig 2. ELSD algorithm flow chart To apply the ELSD algorithm in the log abut detection the following changes were needed: Discard detection of linear segments which is unnecessary for the given task. Discard elliptical and circular arcs with the minimum curvature of the line, since these contours describe linear fragments. Discard contours with anomalously minimal and maximum radii after analyzing the radiuses of the arcs. Generally the log abuts in the image have approximately equal radius, and any significant deviation from this range indicates obvious foreign objects in the image. Discard segments of curves with small lengths after analyzing the lengths of the arcs as far as these arcs generally belong to the background objects (grass, sawdust). Major arcs are completed to a full ellipse or circle.
Result of the automatic detection is shown in Fig. 3.
a
b Fig. 3 – Result of the automatic detection c
Detected abuts are highlighted with color marker. The volume measurement is performed
according to the requirements document (GOST 32594-2013 and GOST 2708-75). The particular
measurement method is selected in the global settings.
2.2. Editing
After the automatic algorithm execution some abuts may be undetected while other objects in the
image may be detected incorrectly, so the manual editing should be implemented. Deleting of the
objects is implemented by selecting “–” tool and clicking objects in the image, whereas the tool “+”
should be selected to add new object. Addition of the objects is implemented on the basis of the Lee
algorithm [
Four neighbor pixels are checked for similarity with given pixel. The pixel similarity test function is following: void checkForSimilarity( Point startPoint, Point checkPoint ){ if ( isPixelExists(checkPoint) ){ if(areColorsSimilar( img.getPixel(startPoint.x, startPoint.y), img.getPixel(checkPoint.x, checkPoint.y) ) ){ mask.put( checkPoint, SIMILAR ); next.add( checkPoint ); } else mask.put( checkPoint, NOT_SIMILAR ); }}
method isPixelExists(checkPoint) checks for image array overrun, mask – structure which stores the checked pixels labeled as included to or excluded from resulted image area.
Pixel color similarity test function is flowing: boolean areColorsSimilar(int c1, int c2){ int differenceR = Abs( c1.red -c2.red ); int differenceG = Abs( c1.green -c2.green ); int differenceB = Abs( c1.blue -c2.blue ); int maxDifference = Max( differenceR, differenceG, differenceB ); return maxDifference < step ? true : false; }
step – algorithm sensitivity. 3. Conclusion It is manufacturing execution system “Plateau” that stores the overall information on the manufacturing processes at the low landing, obtained from various sources including the mentioned software. It allows the key indicators of the forest enterprise performance to be analyzed for any period, planning the further management strategy. MES system enables 24/7 access from anywhere through its implementation over the web. Self-service concept allows the management and the stuff to get all information required without any prior training and special skills.
The software “FoRest” operating principle involves the automatic detection of the logs’ cross-cut ends of the image, calculating the diameters of each cut using calibration coefficients and, finally, cubic capacity of the measured pile based on the obtained data and prior information about the average length of the log pile. The result of the measurement is formed as a report and copied into the data store of the analysis system with following tags: type of operation, raw product (tree trunk/wood assortment), timber species, connected staff (crew, driver, tallyman). Approbation of the software was performed in the logging enterprise under the manufacturing conditions. According to the testing results the average error for the log pile photogrammetry measurement is of 5.14% with maximum error of 9.2% in comparison with manual measurement. Industry standards establish the maximum volume measurement error for the round timber at the level of ±12%. Thus, a method of the log piles photogrammetry measurement using the developed algorithm can be successfully applied in the activity of forest enterprises. “FoRest” is successfully used at the forest enterprises of the Ural Federal District at the moment databases with noise. Proceedings of the Second International Conference on Knowledge Discovery and Data Mining (KDD-96). AAAI Press. 13. Landau U.M. Estimation of a circular arc center and its radius // Computer Vision, Graphics and Image Proc, 1987, Vol. 38, p. 317–326.