Research on Explosive Identification and Localization for Explosive Filling Robot

At present, in the domain of blasting engineering, the predominant approach for filling explosives involves manual labor. To facilitate the smooth grasping of explosives by the robot, a visual detection algorithm is proposed. This algorithm is designed to identify and locate explosives, enabling the robot to replace the manual process of loading explosives in demolition blasting scenes, an experimental platform for explosive identification is built to simulate the demolition blasting environment, and machine vision and YOLOv5 neural network are used to identify and locate explosives. Firstly, based on YOLOv5 convolution neural network, the identification experiment of explosives under different environmental factors is carried out, and the location of explosives is located based on the principle of binocular vision measurement. Finally, the pose and size of explosives images processed by image morphology are detected. The depth and pose information of explosives in three-dimensional space are obtained to provide the information needed for explosive loading robot to successfully grasp explosives. The experimental results show that only when the camera is within 800mm from the explosive, the shielding degree of the explosive should not exceed 50%, and the dust concentration is lower than 100g/m3, can the position and attitude in the explosive space be accurately detected.