An Improved YOLOv5 Algorithm for Apple Object Recognition and Localization

In order to address the low apple recognition rate and inaccurate positioning in complex orchard environments, a method to improve the accuracy of the detection algorithm was studied, and an improved YOLOv5-based apple object detection method was proposed. By obtaining the internal parameter matrices of the left and right cameras of a binocular camera, matching the pixel points of the left and right images, and using disparity mapping, the two-dimensional coordinates of the apples were transformed into three-dimensional coordinates, thereby obtaining the three-dimensional coordinates of the apples. Additionally, considering the issues of overlapping and occlusion of apples in complex orchards, the feature fusion network in the original YOLOv5 algorithm was improved to enhance recognition efficiency and accuracy. Experimental results demonstrate that this improved algorithm effectively achieves apple recognition and positioning in complex environments.