Research And Application Of UAV Intelligent Prospecting System

In geological prospecting and prospecting work,many areas often face many difficulties in data collection due to factors such as terrain and environment,and it is impossible to grasp the high-precision aeromagnetic and high-resolution low-altitude image data of the research area.Existing aeromagnetic surveys have disadvantages such as difficult operation,difficult deployment,and high cost,and cannot meet the needs of rapid surveys.The resolution of traditional high-resolution satellite remote sensing images is usually at the meter level,and the prospecting information obtained for mineral prospecting and exploration is limited.Therefore,the current prospecting and exploration work is in urgent need of a fast data collection equipment.In this paper,a rotor UAV rapid prospecting system is designed.By acquiring low-altitude aeromagnetic and aerial images,and applying deep learning algorithms for real-time analysis,it can realize rapid prospecting and exploration tasks in complex environments.In order to realize fast prospecting technologies such as rapid collection of low-altitude high-precision aeromagnetic data and real-time detection and identification of aerial images.The development and testing of the aeromagnetic measurement system of the rotor UAV and the research on the identification algorithm of the ore prospecting mark in the aerial image were carried out.Based on the hardware and software systems of the DJI M300 RTK UAV flight platform and the DJI Payload SDK design system,a UAV low-altitude aeromagnetic acquisition and recording system was developed.Aiming at the instrument measurement error of the three-axis magnetic sensor,the ellipsoid fitting method is used to correct the error,and the attitude of the aeromagnetic three-component vector data is corrected by the coordinate system transformation method,and the three-component aeromagnetic vector data under the geodetic coordinates are obtained.In the research of aerial image recognition,a training data set of prospecting target detection based on UAV images was collected and produced.The YOLOv5 model has been improved,and the convolutional attention module and the adaptive spatial feature fusion module have been added to enhance the ability of small target detection.The improved yolov5 model was trained with a self-made data set,and its m AP(average recognition accuracy)reached 0.75.The algorithm performed well in the recognition of ore prospecting targets.The UAV rapid prospecting system was tested in the Fangniugou mining area in Yitong County,Jilin Province,and the regional UAV remote sensing images and aeromagnetic anomaly data were collected.The test results show that the improved YOLOv5 model successfully identified the limonite prospecting signs on the aerial images taken by the drone.The system realizes rapid delineation of the prospecting target area and the result of aeromagnetic anomaly interpretation conforms to the distribution characteristics of the ore body.The system improves the efficiency of ore prospecting under complex environmental conditions,reduces the cost of geological prospecting and exploration,and provides new equipment for geological research.