Research on intelligent supervision and safety efficiency enhancement of open pit coal mine blasting based on multimodal sensing of UAV

Blasting is the most dangerous and labour-intensive production process in open pit mines, and it is also the most important part of open pit mining. The rough expansion of production scale makes the traditional manual inspection in the blasting process face great challenges. The traditional manual inspection mode not only faces problems such as overly large positioning of blast holes (greater than 27.4%), low efficiency of filling quality monitoring, and poor detection rate of qualified blasting block size (60% - 72%) in the blasting process, but also encounters problems such as the opacity of three-dimensional geological information in the blasting area, low efficiency of equipment collaborative control and management, and tensile damage to the step structure surface caused by frequent blasting operations, and derivative risks such as the difficulty in early warning of potential safety hazards like fires and landslides. Facing the development of new business forms of low-altitude economy and the intelligent transformation requirements of open-pit coal mines, the technological research and development of a collaborative dynamic supervision system based on unmanned aerial vehicle (UAV) assistance was proposed, and a real-time interlinked control framework of “air-ground-terminal-environment” was studied and constructed: relying on multi-rotor unmanned aerial vehicles in the air, equipped with high-resolution multispectral cameras, infrared thermal imagers, gas monitoring devices, etc., high-precision three-dimensional real-scene modeling of the blasting surfaces of 170 blast holes can be efficiently completed; a multi-parameter sensor network is deployed at the ground layer to construct a blasting vibration monitoring network with a depth of up to 10 meters; the terminal layer has developed a dual-modal target detection model and an improved YOLOv8 m-MSFA visual algorithm. By introducing the multi-scale feature attention mechanism (MSFA), the target recognition accuracy has been increased to more than 96%, and a data-driven dynamic evaluation system for blasting effects has been established. The application results show that: this technical system significantly shortens the parameter iteration cycle dominated by traditional manual experience, compressing it from 6-8 hours to within 2 hours. The iteration efficiency increases by 56.7% - 63.8%, the operation efficiency increases by 30% - 40%, and the accuracy rate of illegal intrusion early warning reaches 96.4%. It has effectively solved the problems such as large safety hazards and serious waste of resources existing in traditional experience-driven blasting operations.