关键词: 果树滴水线, 施肥, 导航路径检测, 图像处理

Abstract: In order to solve the problem that it is difficult to detect the visual navigation path of unmanned equipment walking around the fruit trees in the process of circular ditching and fertilization in the orchard, this paper proposes an image processing based navigation path detection method of fruit tree drip line. In this method, a CCD camera arranged vertically on the ground is used to collect the projection image of the fruit tree canopy, and the recognition of the vertical projection contour of the fruit tree canopy along the ground and the smooth processing of the drip line are realized, and then the circular walking path of the unfertilized equipment along the fruit tree is determined. In this paper, camera internal parameters and distortion parameters are obtained through camera calibration, and the original image is corrected. Through the quantitative analysis of the distribution characteristics of the image in RGB color space, the mean value method is used to process the image gray. Through the distribution characteristics of the original image and RGB, the environment and tree canopy projection are obviously different in gray value, and the threshold method is used for binary segmentation. Since there are a lot of noise and gaps in binary images, morphological expansion operation is used to fill gaps to highlight the edge contour of canopy projection. Boundary tracking algorithm is used to extract crown contour edges. Beseel curve interpolation fitting method is introduced to delete some inflection points and smooth the contour edges. By comparing the secondorder, thirdorder, fourthorder and fifthorder fitting results, it is concluded that the thirdorder and fourthorder Beseel fitting results are more suitable for navigation path requirements. Fixed in one place, the camera shot in sunny and cloudy conditions respectively collectes image, extract drip line navigation path, respectively, the use of thirdorder and fourthorder Beseel curve fitting sunny and cloudy image edge contour, using the fourth order fitting results more in line with the actual requirements, the average error is 19.5 pixel, pixels on average relative error is 2.6%, The average image processing speed of each frame is 27 ms, which can better meet the requirements of navigation accuracy and realtime. This study can provide reference for automatic navigation of fertilization platform along drip line.

Key words: fruit tree drip line, fertilization, navigation path detection, the image processing