Research On Multi-object Spatial Position Measurement Technology Based On Vision

In recent years,moving target detection and positioning technology has been widely used in the fields of national defense,weapon testing,scientific research,Unmanned Aerial Vehicle（UAV）target tracking and traffic monitoring vehicle positioning.Different from the application of it in other fields,the complexity of the weapon test ranging environment and the particularity of the spatial distribution of multi-target projectiles make target detection and positioning extremely difficult.In order to meet the needs of real-time accurate positioning of multi-target projectiles on ranging,vision-based testing and tracking technology was studied by this thesis.The double-station cameras are used to collect multi-target image sequences.According to target detection and multi-target tracking technology,combined with the epipolar geometric matching algorithm,the accurate tracking and positioning of multi-target projectiles in the complex environment of the shooting range is realized.The main contents of this research are as follows:（1）The overall scheme of projectile multi-target positioning system was designed.The composition and function of the test system were introduced,the selection of cameras and lenses was completed,the multi-target spatial position measurement model was established.The design methods such as the layout of the camera installation position,object detection,multi-target tracking positioning and stereoscopic matching were studied.（2）The spatial multi-object detection algorithm was researched.Aiming at the complex environment in the field test,an improved Vi Be algorithm was proposed based on Visual Background Extractor（Vi Be）target detection algorithm.The foreground targets were segmented accurately,with segmentation threshold which was adjusted according to the complexity of the background to updata the background model in real time.The false alarm target was eliminated by judging the similarity of the edge gradient feature model between the foreground area and the moving area.The experimental results show that the algorithm has strong robustness to the complex environmental factors,improves the integrity and accuracy of prospect target detection,and is suitable for the field test environment of dynamic targets.（3）The spatial multi-target positioning and stereo matching algorithm were researched.According to the target frame cells composed of space-time features,a algorithm that establishes the spatiotemporal feature information association matrix of multiple targets between adjacent frames was proposed to solve the problem of targets remarked caused by the particularity of the spatial distribution of multi-target projectiles and the difference in the explosion time between the projectiles in tracking and position of multi-target.The experimental results show that the algorithm in this paper can effectively solve the problems of overlapping occlusions between targets,as well as new targets and disappearance of original targets,and achieve accurate tracking of multiple targets.Finally,according to the characteristics of projectile imaging and the commonly used stereo matching recognition algorithms,the epipolar geometry algorithm was selected to achieve the correct matching between the left and right view multi-targets.（4）By setting up an experimental test platform and simulating multi-objective projectile motion experiments,the effectiveness of the positioning algorithm in this paper was verified.Experimental results show that the measurement scheme proposed in this paper can achieve accurate measurement of multi-target spatial position,and the positioning accuracy is better than10 mm,which meets the needs of multi-projectile target positioning and has strong adaptability to the environment.