Design of a Visual Detection Method for Prohibited Item Packages Detection Based on the YOLO Model

In this paper, we introduce a prohibited item package detection system based on the YOLO model. Utilizing the state-of-the-art YOLO as its core object detection model, with the integration of residual connections and deep attention mechanism, the detection accuracy for prohibited item in the test dataset can exceed 85%. The model has a certain anti-interference ability against item occlusions and X-ray imaging distortions. It can efficiently recognize and detect prohibited item in a large volume of postal item image data, and it achieves superior detection results for overlapping and occluded prohibited items. The prohibited item package detection system based on YOLO accurately identifies prohibited items, offering an effective technical solution for the postal monitoring of prohibited items.