A Study on Weed Detection Based on Machine Vision
(0) Weed is one of the important factors ，which damages the quality of the agricultural products. At present the method to weed is to spray the herbicide and the means to spray is the well-distributed spraying. This method not only improves the cost of agriculture but also damages the quality of field and pollutes air. The study on infield weed detection using machine vision is significant to reduce herbicide dosage and environment pollution by applying the variable-rate spraying. The paper studied the method to detect the inter-row weed based on image processing technology. Firstly, applying color feature increased the contrast between green plants and soil, separated the plants from complex background, changed real color image to gray-level image and pre- processed the image to remove the noise. Secondly, some algorithms of image threshold segmentation were researched and selected a suitable segmentation method, namely changing threshold segmentation. The algorithms could transfer gray-level image into binary image with stabilization and no distortion. Finally, we studied the BPNN (Backpropagation NN) which was used to detect the weed using the functions provided by MATLAB, and had many training and imitation. The results of the experiment showed that in order to improve the rate of identified the weed properly and built up the stable foundation of the further research, the training times and the rate of sample learning and identifying should be considered in the net structure.
image processing weed detection machine vision
Li Dong-ming Wu Bao-zhong Liu Ya-ju Ren Zhen-hui Sun Yu-mei Du Bo
Mechanical and Electrical College of Hebei Agriculture University of China, Baoding, 071001 Foreign Languages College of Hebei Agriculture University of China, Baoding, 071001
国际会议
北京
英文
2007-08-05（万方平台首次上网日期，不代表论文的发表时间）