"Prevention of" Dangerous "Du Jian-Intelligent Inspection Robot for Deep Underground Roadway Working Conditions" is a multi-functional and integrated inspection equipment with surrounding rock monitoring as the core suitable for mine roadway engineering. Computer vision is used to monitor surrounding rock deformation, which greatly solves the problems of complicated manual measurement process, large amount of layout work and high installation cost of the existing surrounding rock deformation measurement system. The movement mechanism of the robot adopts crawler structure, it has good obstacle crossing ability and ground adaptability. With laser SLAM and visual SLAM, it can realize the positioning, navigation and obstacle avoidance functions of the robot in underground space. Compared with the single motion path of the track sliding inspection robot, it has higher degree of freedom and more flexibility, and can carry out inspection operations in more complex special environments such as mine tunnels. Equipped with a variety of sensors, it can monitor and warn the environmental parameters such as toxic and harmful gas concentration and temperature in the roadway. The robot has the functions of monitoring the surrounding rock deformation of the roadway, building maps of underground space and automatic navigation inspection, and carrying sensor modules to collect on-site environmental parameters in real time. In addition to the inspection work of mine tunnels, the robot can also complete the inspection work under some scenes such as underground tunnels and underground pipe corridors. 1. Product function This product is a multi-functional investigation and early warning robot with roadway surrounding rock monitoring as the core. It is suitable for the inspection work of mine shaft and roadway. It has many functions, such as underground space map positioning, roadway surrounding rock deformation monitoring, dangerous gas monitoring, dust monitoring, fire warning, obstacle crossing and so on. The deformation monitoring function of roadway surrounding rock uses computer vision image processing method and feature extraction method to monitor the deformation of underground roadway roof and floor and two sides of surrounding rock. The SLAM technology is applied in the robot to realize the function of positioning navigation and obstacle detection, the robot can draw the map of the interior of the roadway in the process of traveling, and visually show the complex spatial conditions in the roadway. The drawing of the roadway map is the basis for the robot to complete the automatic inspection. At the same time, a variety of dangerous gas sensors carried by the robot can monitor the concentration of dangerous gases such as gas and carbon monoxide in real time, issue warnings in areas with abnormal dangerous gas concentrations, and mark them on the map to issue early warnings. Equipped with infrared camera to collect temperature visual information, detect the heat source in the roadway and prevent mine fire. The robot adopts wireless network signal control, and can realize off-line automatic navigation after drawing, which can replace manual completion of underground roadway surrounding rock deformation detection and dangerous environment reconnaissance and early warning, minimize labor costs, reduce safety risks, and escort mine safety production. 2. Scientific product About SLAM synchronous map building navigation technology: SLAM is a collective term for synchronous mapping and autonomous positioning and navigation technologies, and it also includes a variety of SLAM methods with different principles. Among them, the full 3-dimensional laser SLAM has high precision and can scan and map the environment in which it is located. However, its disadvantages are large weight and volume, large amount of laser data transmitted back, and high cost (the price ranges from 3 to 100000). In recent years, it is mostly used for scientific research in the field of underground space. Compared with the full three-dimensional laser SLAM, the three-dimensional visual SLAM is not as accurate as the latter, but it can also reflect the three-dimensional environment in the underground space more accurately. Combined with 2-dimensional laser SLAM, it can meet the needs of the robot's underground inspection work. About visual deformation measurement of surrounding rock: The use of visual images to detect the deformation of surrounding rock is not the first time, the 2020 article "underground roadway surrounding rock deformation rapid and accurate monitoring system and its monitoring methods and production process" proposed a fixed camera to do a good contour marking of the roadway deformation detection device. In this method, each roadway deformation acquisition point needs to be laid a camera and need to lay communication cables, so the resource consumption is large, and the pre-system installation is also very cumbersome. The visual surrounding rock detection function of the underground space operation environment investigation and early warning robot is realized by a modular device, which is different from the independent visual surrounding rock deformation monitoring system. The module is mainly used as the function expansion module of the inspection robot. By combining with laser SLAM technology, the precise positioning of the camera (mounted on the robot) in the roadway can be realized, avoiding obvious errors caused by different parking points and angles of the robot. In contrast, this product only needs one camera to complete the collection task of all roadway deformation collection points, and the preliminary preparation work only needs to be marked on the roadway wall, without hanging lines and laying lines, which greatly reduces the investment of resources and manpower, and can put the detection equipment into use faster. Because they are all through the image acquisition of the roadway section after marking the contour, the calculation principle of the roadway deformation is basically the same, so the measurement accuracy is equivalent, which is 0.01m. 3. Innovations (1) Based on SLAM real-time positioning and mapping, SLAM and VSLAM realize the functions of positioning, navigation and monitoring spatial changes, and can intuitively show the internal form of space. Compared with the limitations of the motion space and route of the orbital inspection robot, its degree of freedom is higher; compared with the full three-dimensional laser SLAM with high precision, high cost, large volume and weight, its cost performance is higher. Laser SLAM is also the guarantee that the camera can accurately locate the position and angle in the roadway. (2) The vision module embedded in the robot collects the roadway contour mark by high-precision camera, and then calculates the deformation data of the roadway through image processing, feature extraction and database comparison. This kind of surrounding rock deformation detection method solves the problem of cumbersome manual measurement process to a large extent. Only one camera is used to complete the acquisition of all roadway deformation points. Compared with the layout of the existing surrounding rock deformation measurement system, the engineering quantity is smaller and the installation cost is higher. The collected roadway contours are stored in the database of the computing platform, and the next contour acquisition of the same section is compared to determine whether the roadway is deformed. The roadway deformation data in the database is also used to optimize the learning of the feature model to improve the accuracy of feature extraction.