Research on improved YOLOv5s algorithm based on optical sensors for AGV object detection

With the continuous development of automation technology, the application of Automated Guided Vehicles (AGVs) in logistics, manufacturing and other fields is becoming increasingly widespread. Target detection, as one of the key technologies for AGV cars to achieve autonomous navigation and intelligent operations, directly affects the overall performance of AGVs in terms of performance and efficiency. Optical sensors (such as cameras), as the core components of environmental perception in intelligent transportation systems, provide rich scene information for AGV and intelligent transportation applications through high-resolution image acquisition and real-time data processing. This paper proposes a target detection algorithm based on improved YOLOv5s, which combines the visual perception ability of optical sensors with the edge computing ability of Aixin Yuanzhi AX620A chip, significantly improving the real-time detection efficiency of the model in complex traffic scenes. By replacing the SiLU activation function in the YOLOv5s model with the ReLU activation function and optimizing the collaborative working mechanism between the model and optical sensors, the experimental results show that the improved model improves computational efficiency while maintaining high detection accuracy. The inference speed reaches 25fps and can accurately identify targets such as pedestrians, two wheeled electric vehicles, and motor vehicles, providing strong technical support for the practical application of AGV in intelligent transportation.