A Recognition Scheme Based on K-means Feature Clustering for Obscured Apple Object

Abstract

At present, the recognition effect of the mature apple automatic picking vision system is not ideal. To solve this problem, this paper proposes a method based on K-means feature clustering for the recognition of the occluded apple object. That is, using K-means algorithm to cluster the image in L*a*b color space, and using the *a component to distinguish the object from the background to improve the segmentation accuracy. Then, the real contour of the segmented object is extracted by using the convex hull algorithm. Finally, the least squares algorithm is used to reconstruct the extracted object contour. The experimental results show that the proposed scheme can accurately segment and recognize the occluded apple object, which lays a solid theoretical foundation for the application of apple automatic picking vision system to achieve precise picking.