Research On The Field Self-Running Control Of Spray Insecticide Machine Based On Machine Vision

Plant protection machinery has a vital role to play in the plant diseases and insect pests forecasting. Although many important achievements have been made in china, it is still hard to meet the needs of agricultural development. Because of the increasing demand for lots of working of mechanics and constant loss of rural labor force drain,therefore, automation, intelligent transformation of plant protection machinery is of great significance to solve the current problems.Based on the above reasons, this article takes the self walking control of the spray insecticide machine as research object. Preprocessing method and the road center position determination method of the field road image were studied from the field walking road image processing and self-running control these two aspects and provided offset data for self-running of spray insecticide machine. At the same time,Experimental study was carried out on the road image segmentation accuracy and the effect of self-running of spray insecticide machinexthe main conclusions are as follows:(1) Image processing primarily includes the choice of color space, gray processing,filtering processing and threshold segmentation. This aspect will directly affect the stability and accuracy of data post-processing. The experimental results showed that: a.color space YCbCr has a good real-time performance, strong adaptability to light and great image segmentation capability to complex road; b. Component method is easy,fast and of great results; c. The results of bilateral filtering is best while the real-time performance is poor; d. Method of maximum classes square error is simple and high efficiency. It has a good country road image segmentation results.(2) Roads central position was located through taking statistics of pixel and Gaussian fitting of image preprocessed. It is the base of deviation judgment and self walking control of spray insecticide machine. Took experiments on common road and complex road and the experimental results run as follows: Maximum error of central position located automatically and human being were 8 pixel and 12 pixel; The average error were 5.1 pixel and 8.9 pixel; Corresponding actual error were 4.85 mm and 8.54mm; The maximum error rate were 2.50% and 3.75%. It shows that this method has a good adaptability and it is more accurate on common roads.(3) Accuracy of country roads image segmentation means the goodness of fit between road segmentation and ideal segmentation. The mean of accuracy ofsegmentation on the two type!s road were 95% and 83%. The accuracy of complex road is obviously lower. However it has little effect on the road image central position localization.(4) To realize country road self-walking, off-set station was firstly judged through the deviation between value of actual road central position and ideal central position.Then front wheel steering was controlled through Simulink to correct the deviation progressively. At last rear wheel was driven to make the machine walking along the country walking road by itself. Experiments showed that this spray insecticide machine can realize self-running under low speed, good road condition and regular road. The maximum error can be kept less than 28 mm.