Development of agricultural implement system based on machine vision and fuzzy control

Abstract

To achieve accurate mechanical inter-rows weeding, an agricultural implement guidance system based on machine vision was designed. The guidance system consists of a color video camera, an industrial panel PC, a lateral displacement controller, a GPS receiver, a hydraulic system, and an agricultural implement. To improve the accuracy and reliability of the guidance system, the choice of color space, the method of guidance line detection, and the method of controlling the implement were investigated. First, considering the adverse effect of illumination variation on image processing, the HIS (hue, saturation, intensity) color model was used to process images, and a threshold algorithm based on the H component was used to produce grayscale images. Second, according to the characteristics of the crop rows in the image, a method of crop line identification based on linear scanning was proposed. To approximate the trend of a crop row in the image to a line, pixels at the bottom and top edges of the image were selected as two endpoints of the line. Candidate lines were created by moving the position of these endpoints. The line with the most target points was regarded as the crop line. Finally, fuzzy control was used to control the agricultural implement. This algorithm can effectively control the agricultural implement tracking the guidance line. Path tracing experiments were conducted at three different speeds of 0.6, 1.0 and 1.4m/s in the corn field on a sunny day. The maximum lateral errors were 4.5cm, 5.5cm and 6.8cm at the three speeds. The average lateral errors were less than 2.7cm for all speeds. The experimental results demonstrated that the guidance system successfully adapted to changes in natural light and had good dynamic performance at all speeds.