Application Of Machine Vision In The Weeding Robot

Farmland weeds overrun, especially in crop seedling weeds grow so easily affect crop in cropseedling stage is crucial for weed control, the current mainstream for chemical herbicidesweeding way, but for environmental damage is bigger, so do not rely on the intelligent chemicalherbicides weeding robot research has received widespread attention.Based on the summary of weeding robot at home and abroad on the basis of research results,through a series of theoretical study and experimental study, developed a set of real timerecognition field thinning grass machine vision system, through the experiment, proved that thesystem can complete the weeds of the field, in the process of weeding no pesticide use and canreach the expected effect of work.First hardware design of the machine vision, including image acquisition device, imageprocessing, equipment selection, installation and debugging software development using c++language, the library is developed based on OpenCV vision can be real-time image recognitionthe field of image processing software. For the field acquisition of image preprocessing, medianfiltering of the RGB color for the image of the field test, green seedlings in the grass crop andsoil background there were significant differences in color features. Based on the color of grasscrop and soil background characteristics difference adopted the method of super greenbinarization image segmentation method, using adaptive threshold method for image thresholdsegmentation. Image post-processing further eliminate noise by means of morphological closingoperation, image edge detection and centroid localization and thus drive robots to makedecisions.Using soil tank lorry and Deere tractors front suspension weeding robot804machine visionrecognition test, test the time consuming of image recognition and speed of robot hand weedingknife, prove that the system can real-time recognition field thinning finish mowing grass.