Orchard Mobile Robot Detection System Based On Combination Information

To achieve orchard operation automation, this study explains the current agricultural vehicle navigation method. By comparing the existing methods of navigation, the problems which is in the current navigation have been discovered. So as to settle the problem of orchards environment information incomplete obtaining, this study proposes a multi-sensor fusion method which is based on machine vision and laser navigation. The main research contents include:(1) In this study, the leopard 4040 tractor was modified into a navigation platform. The navigation system, depended on multi-sensor fusion method, has been established. the use of sensors and the hardware was introduced.(2) This study developed a navigation platform in orchard, which is based on laser rangefinder. A kind of interference elimination methods of orchard fruit trees location identification method in adjacent fruit trees rows is proposed. According to the location information of fruit trees, the orchard navigation path was fitted. A fuzzy controller was designed with using the course angle error and the lateral deviation error as the input. The angle of front front-wheel turning was used as a output of the fuzzy controller. The navigation system was tested with the tractor travelling 30 m at the speeds of 0.27m/s,0.33m/s or 0.45m/s. In simulated navigation environments that involved double fruit tree rows, the maximum lateral deviation of the navigation system at the three speeds were0.077 m, 0.113 m and 0.103 m. In simulated navigation environments that involved single individual fruit tree rows standing to the left of the tractor, the maximum lateral deviation of the navigation system at the three speeds were 0.04 m, 0.09 m and 0.088 m. In simulated navigation environments that involved single individual fruit tree rows standing to the right of the tractor, the maximum lateral deviation of the navigation system at the three speed were 0.056 m, 0.067 m and 0.08 m. In simulated navigation environments that involved double fruit tree rows with some trees lost, the maximum lateral deviation of the navigation system at 0.33m/s were 0.18 m. The study revealed that it was feasible to use a navigation system to navigate orchard machinery.(3) Digital image processing technologies are used to handle the orchard image forprocessing that was obtained by machine vision. Firstly, the images was transformed into HSV space. The OTSU method was used to proceed image threshold segmentation. Finally the fruit tree-crown outlines are obtained.(4) The fruit tree canopy navigation line are extracted by using Hough transform from the processed image, and the orchard navigation line was fitted. Through the geometric relation between image space coordinate system and world geodetic coordinate, the course angle conversion formula was deduced.(5) The kalman filter was used to fuse the course angles form machine vision and laser range finder. The orchard multidimensional combination navigation information fusion method was obtained.