Research On Automatic Interpretation And Pose Calculation For Shooting Range Target

Optical measurements are an important means of external measurement for space launch and shooting-range test identification,providing important data on external ballistics,target characteristics,etc.for weapon performance assessment and feature event analysis.With the increase in the types of shooting-range test tasks,the increase in the testing requirements of new types of weapons,and the improvement of the capacity of shooting-range optical measurement equipment,the shooting-range optical measurement image interpretation processing urgently needs to improve the processing efficiency and refinement level,requiring efficient,high-precision and highly applicable target detection and interpretation algorithms and high-precision solving methods to build efficient distributed real-time interpretation system.In this paper,we focus on the difficult problem of optical image processing in the shooting-range.1.A dual-categorical deep learning target detection algorithm with better adaptation to target types and characteristics is proposed.To address the problem of poor adaptability of traditional target detection algorithms due to the large number of target images and target types and changes in target characteristics,a dual classification deep learning target detection method is proposed.The method is based on Yolo V3,a deep learning target detection framework,which improves the single-attribute classification to dual-attribute classification in the network output layer based on the dual-attribute characteristics of the brightness and shape of the target of the target of the target range;implements automatic labeling of the sample based on the growth of the target area;and increases the detection result confirmation link using the sequence image target constraint.The actual image training and detection results under multiple scenarios of the target field showed that the proposed method achieved 99% detection success rate and 72% average positioning accuracy of IOU.2.A linear measurement method for the target axis with minimal residuals in the directional image angle of the anchor point is proposed.For the measurement problem of the target axis,a linear algorithm is proposed to take a feature point on the target axis as the anchor point,and the direction of the line passing over the point as the anchor point direction.The algorithm first uses spatial intersection to obtain the spatial position of the anchor point,minimizes the angle between the projection of the anchor point direction on the image plane and the target two-dimensional axis as an objective function,and replaces the equivalence of the angle expression that cannot be solved linearly with a simple linear expression,thus solving the axis direction linearly.The simulation results show that the proposed linear algorithm achieves accuracy consistent with the iterative algorithm and still gives high accuracy results when the intersection conditions are poor.3.A method for automatic interpretation of large targets and high-precision attitude resolution under close observation conditions is proposed.Drift and attitude are important parameters for the initial launch segment of a rocket or missile and require close observation,with problems such as sequential entry and exit of cooperative markers on the target,target deformation and blurred imaging motion.In order to solve the problem that the interpretation of the target cooperative flags in and out of the image requires a lot of manual operation and is prone to errors,an efficient automatic detection interpretation method for cooperative flags combined with model prediction and template matching is proposed;in order to solve the problem that the target deformation leads to the reduction of the position solution,a method for estimating the deformation and attitude synchronization by integrating the deformation model of the target is proposed;in order to solve the problem that the motion blurring leads to the reduction of the position solution,an iterative position solution combining the target motion speed and camera parameters is proposed to correct the position of the image point.The simulation results prove the validity of the interpretive method.The simulation of the actual scene conditions verified that the deformation error is less than1 mm.The proposed solution method is significantly improved relative to the method that does not consider motion blurring,and the solution error is less affected by the target’s speed.4.Research and implementation of a distributed real-time interpretation and solving system for shooting-range.In response to the need for distributed real-time interpretation of target-site images and in combination with the actual characteristics of the shooting range,the optimal filtering is used to predict the integrated angle,give the target’s predicted position,and use nuclear-related filtering for tracking,which has achieved very good tracking results.The key software technologies in the distributed real-time interpretation and solving system of shooting-range images are studied,including the design technologies of system distribution,timeliness and robustness,and a distributed real-time interpretation and solving system of shooting-range images is designed and implemented.The simulation results show that the improved real-time interpretive method has high tracking accuracy and timeliness to meet the real-time requirements.The proposed algorithm and interpretation software are applied to the actual task of the shooting-range,which improves the efficiency of the target range image processing and interpretation accuracy,and plays an active role in the technical progress and level improvement of the shooting-range image interpretation measurement.