Approaches to Address Occlusion and Small Target Challenges in Restaurant Rat Detection

In this study, an improved target detection algorithm is proposed for the problem of rat detection in restaurant environments. The restaurant environment is characterized by obvious light changes, complex occlusion, and ground reflections, which bring more challenges to the accurate detection of rats. For this reason, we make the following improvements to the YOLOv5 algorithm. First, we introduce the Glare Suppression Module, which utilizes an adaptive feature correction method to reduce the interference of highly reflective areas such as oil and water stains on the ground on the detection model, and to improve the recognition ability of the rat target in low-contrast scenes. Secondly, combined with the Background Adaptive Modeling (BAM) technology, the stability and generalization ability of the model is improved by dynamically learning the background features of the restaurant environment to reduce the false detections caused by the activities of the restaurant personnel, changes in the tables and chairs, and other factors. The experimental results show that the improved YOLOv5 algorithm can effectively improve the precision and recall rate of rat detection, reduce the false detection rate, and ensure the real-time performance in the restaurant environment, which provides efficient and reliable technical support for restaurant hygiene monitoring and pest rat prevention and control.