Research on Intelligent Sorting and Handling System Based on Lightweight Improvement Algorithm

To address the challenges of high computational load in algorithms, difficult deployment on embedded devices, and insufficient low-temperature environmental adaptability in intelligent sorting systems for cold chain logistics scenarios, this paper proposes a multi-sensor fusion architecture, the proposed system integrates visual perception, robotic arm control, and mobile robot collaboration functionalities. By establishing a lightweight algorithmic improvement scheme in visual perception, enhances the overall performance and efficiency of the system. At the algorithmic optimization level, lightweight enhancements are implemented based on the YOLOv5 framework: 1) A depthwise separable convolution reconstructed backbone network is employed to reduce parameter count; 2) An enhanced Focal-EIoU loss function is incorporated to improve detection accuracy while reducing motion control computational complexity. Experimental results demonstrate that the improved algorithm significantly enhances sorting accuracy and reduces processor workload.