盲道识别与障碍物检测的多任务模型OA

This paper proposes the Amaterasu-YOLO multi-task model aimed at improving the accuracy and efficiency of blind path area segmentation and obstacle detection.The model integrates an Adaptive Cascade Module(ECD)and a Multi-Receptive Spatial Attention Module(MRSA),enabling high-precision blind path segmentation and obstacle detection in complex urban en-vironments.By leveraging multi-task learning,Amaterasu-YOLO not only optimizes the joint tasks of blind path segmentation and obstacle detection but also significantly reduces computational burden,enhancing the model's efficiency on resource-constrained edge devices.Experimental results show that Amaterasu-YOLO achieves good performance in both blind path seg-mentation and obstacle detection tasks,with segmentation accuracy reaching 90%and obstacle detection accuracy reaching 85%.Compared to traditional single-task methods,the model demonstrates stronger robustness and practicality,with broad ap-plication potential in smart city development and ensuring the safety of visually impaired individuals.