Research on Automatic Detection of Screw Fastening Positions for Robotic Arms Based on Visual Deep Learning

Screw fastening is a critical process in industrial automation assembly, which impose stringent requirements on the accuracy and real-time performance of detection systems. This paper proposes an automated robotic arm screw position detection system based on deep learning. To address challenges such as lighting variations and background interference in complex industrial environment, a high-resolution image acquisition module is designed, coupled with an adaptive histogram equalization algorithm to enhance image quality. A multi-scale object detection model, based on the VGG16 backbone network, achieves precise screw position recognition and localization through efficient feature extraction. Furthermore, a hardware-software collaborative feedback control mechanism is designed to enable real-time linkage between detection results and robotic arm operations. Experimental results demonstrate that the system achieves a detection accuracy of 99.8 % on an industrial scene dataset, with a single-frame detection time of 32 milliseconds, meeting the demands for both real-time performance and high accuracy. The research findings indicate that the proposed method exhibits excellent robustness and reliability, providing technical support for industrial automation assembly and offering a reference for the application of deep learning in industrial scenarios.