AN OBSTACLE DETECTION SYSTEM BASED ON MONOCULAR VISION FOR APPLE ORCHARD ROBOT

Obstacle detection is one of the key problems in the autonomous movement of intelligent mobile robot. This paper focuses on the study of obstacle detection in the complex apple orchard. First, using L* a* b* (colour space for non-self-illuminated, named by CIE 1976 L* a* b* or CIELAB) colour model to represent the apple orchard image, the obstacles in the robot's walking area can be effectively segmented from the complex background by one-dimensional fuzzy entropy image segmentation algorithm. Then, the coordinate of the point close to the robot of the obstacle is obtained. Finally, the distance between the robot and the obstacle is estimated using the inverse perspective transformation. The experimental results show that the maximum absolute error is 7.87 cm and the maximum relative error is 6.53%. The results can meet the accuracy requirements of the intelligent mobile application for an apple orchard robot.