For wheat harvester working path planning under complicated boundary problems, because the agricultural machinery operation paths largely determine the operation quality, efficiency, and cost, the cutter path planning model based on operation direction was set up, and an algorithm was presented in view of the convex and concave side boundary farmland with the least number of turns. The algorithm included the scanline method to calculate the number of turns with convex edges and the imaginaryregionfilling method to calculate the number of turns with concave edges. The rotation step method is used to calculate the optimal working direction corresponding to the minimum number of turns, and the optimal path trajectory of the harvester is obtained. The results of the simulation field boundary show that the optimization algorithm can effectively obtain the operation direction and path trajectory with the least number of turns. In the simulation case of the convex edge field, the optimization degree of the optimal operation direction can reach 11.32% compared with the path along the long side direction. In the field simulation case with concave edges, the optimization degree of the optimal operation direction is 5.66% compared with the path along the long side. The path planning model and algorithm can provide a certain decision basis for harvester harvesting in an actual complex boundary environment.

The identification of wheat heading at flowering stage can be used to guide the later water and fertilizer management, disease control, yield prediction and other aspects. In order to realize accurate and automatic ear counting, this paper proposes a new ear counting method based on color features. The color of wheat ear is very close to that of leaf and stem at heading flowering stage, and the common color features can not segment wheat ear effectively. In this paper, we extracted wheat ear effectively by color histogram equalization and red green normalized difference index. Aiming at the problem of wheat spike adhesion in the image, this paper uses the improved Harris corner detection algorithm to verify the wheat images taken at vertical angle and 45° angle respectively. Through the sample image counting experiment, the accuracy is 96.06% and 94.74% respectively. The results showed that there were obvious color differences in wheat ears, leaves and stems after equalization, and the color features could be used to extract the images of wheat ears at flowering stage in field environment. The detection of skeleton intersection point after wheat ear thinning can accurately count the conglutinated wheat ears with high counting accuracy, which can be used to reflect the situation of wheat heading in this period.