Multi-sensor data fusion and navigation line extraction method based on discrete factor

Objective  In view of the fuzzy and irregular shape of the path boundary in the corn field, the common field navigation line extraction algorithm will have the problem of excessive deviation in practical application of agricultural robot. In this paper, a navigation line extraction algorithm based on discrete factor fusion of camera and 3D LiDAR is proposed for the field of corn at 3rd−5th leaf stage.

Method  First, three-dimensional lidar was used to obtain corn plant point cloud data. At the same time, the green feature binary images were obtained from the images taken by the camera using the super-green algorithm and the maximum between-cluster variance method, and then the point cloud data after cluster analysis were projected onto the target bounding box in the image. A multi-sensor data fusion support model was constructed for feature recognition. Finally the acquired feature center point was fitted as the navigation baseline.

Result  The algorithm could adapt well to complex environments and had strong anti-interference ability. The average processing time of a single frame was only 95.62 ms, and the accuracy rate was as high as 95.33%.

Conclusion  The algorithm solves the problems of shifting in finding feature centroid and unreliable recognition results in traditional algorithms, and provides a reliable and real-time navigation path for the robot to walk in corn field.