A deep learning based hazardous materials (HAZMAT) sign detection robot with restricted computational resources

One of the most challenging and non-trivial tasks in robot-based rescue operations is Hazardous Materials (HAZMAT) sign detection dangerous operation fields, order to prevent further unexpected disasters. Each HAZMAT has a specific meaning that robot should detect interpret take safe action, accordingly. Accurate real-time processing are two important factors such robotics applications. Furthermore, cope with some secondary challenges as image distortion restricted CPU computational resources, embedded robot. In this research, we propose CNN-Based pipeline called DeepHAZMAT for segmentation four steps: (1) Input data volume optimisation before feeding into CNN network, (2) Application YOLO-based structure collect required visual information from hazardous areas, (3) separation background using adaptive GrabCut technique, (4) Post-processing morphological operators convex hull algorithms. spite utilisation very limited memory experimental results show proposed method successfully maintained better performance terms detection-speed detection-accuracy, compared classical modern state-of-the-art methods.