 The Apple Target Recognition Algorithm is a core technology of an Apple-picking robot, but most existing detection algorithms cannot distinguish between occluded apples. A lightweight Apple Target detection method was proposed using improved YOLO V5S, which includes improvements to the bottleneck CSP module, SE module, bonding fusion mode, and initial anchor box size. The experimental results showed that the graspable and ungraspable apples could be effectively identified with high recall, precision, map, and F1 scores. Compared to other models, the proposed improved YOLO V5S model increased its map by 5.05% to 6.75%, compressed its size by 9.29% to 15.3%, and had faster recognition speeds per image. The proposed method can provide technical support for real-time accurate detection of multiple fruit targets for the Apple-picking robot. This article was authored by Bin Yan, Pan Fan, Xiaoyang Lei, and others. We are article.tv, links in the description below.