Harvesting Route Detection and Crop Height Estimation Methods for Lodged Farmland Based on AdaBoost

Addressing the challenge of the current harvester route detection method's reduced robustness within lodging-affected farmland environments and its limited perception of crop lodging, this paper proposes a harvesting operation image segmentation method based on SLIC superpixel segmentation and the AdaBoost ensemble learning algorithm. This segmentation enables two essential tasks. Firstly, the RANSAC algorithm is employed to extract the harvester's operational route through straight-line fitting from the segmented image. Secondly, the method utilizes a 3D point cloud generated by binocular vision, combined with IMU information for attitude correction, to estimate the height of the harvested crop in front of the harvester. Experimental results demonstrate the effectiveness of this method in successfully segmenting the harvested and unharvested areas of the farmland. The average angle error for the detected harvesting route is approximately 1.97 degrees, and the average error for crop height detection in the unharvested area is around 0.054 m. Moreover, the algorithm exhibits a total running time of approximately 437 ms. The innovation of this paper lies in its simultaneous implementation of two distinct perception tasks, leveraging the same image segmentation results. This approach offers a robust and effective solution for addressing both route detection and crop height estimation challenges within lodging-affected farmland during harvesting operations.

Automated summary

This paper proposes a harvesting operation image segmentation method based on SLIC superpixel segmentation and the AdaBoost ensemble learning algorithm, which addresses the challenges of reduced robustness and limited perception of crop lodging in the current harvester route detection method. Experimental results demonstrate the effectiveness of this method in successfully segmenting the harvested and unharvested areas of the farmland.