Research And Implementation Of Intelligent Target Reporting System Based On Object Detection

The construction of digital shooting range is an inevitable requirement for the transformation of traditional military training.As an important component of building digital shooting range,intelligent target reporting system has become a research hotspot.The emergence of intelligent target reporting system has fundamentally solved the disadvantages of manual target reporting,but most of them have the disadvantages of poor performance,high development and maintenance costs,heavy equipment and difficult system calibration.Aiming at the above shortcomings,this paper designs an intelligent target reporting system based on target detection,which mainly consists of three parts:hardware system,target reporting algorithm and target reporting software.1.In order to meet the advantages of low development and maintenance costs and portability,the hardware system is only composed of target reporting terminal,shooting terminal and monitoring terminal.In the selection of hardware materials,in view of the problem that the vision-based intelligent target reporting system cannot be applied to target papers with different color coatings,this paper chooses grayscale industrial cameras as image acquisition equipment.In order to make the system development cost low and have good performance,this paper chooses NVIDIA Jetson Nano as the edge computing device,and selects Intel 8265NGW wireless network card as the wireless communication module to ensure the data transmission speed of the system.2.The target reporting algorithm is mainly composed of target image preprocessing module,target detection and tracking module,bullet hole ring value determination module.(1)In the target image preprocessing module,in order to solve the problems of false target,missed target and waste of computing resources caused by low quality image,a fuzzy detection unit is added.In the unit,aiming at the problem that the ambiguity calculation of the Laplacian operator is interfered by the background,a multi-scale template matching algorithm is proposed to roughly cut the target image to eliminate part of the background interference,so as to determine the fuzzy detection threshold.The accuracy of fuzzy detection is 98.77%.In order to ensure the accuracy of chest ring target effective region extraction,a chest ring target image segmentation method based on Hough gradient circle detection and cumulative probability Hoff line detection is proposed.Comparative experiments show that this method has high robustness and good performance.(2)In the target detection and tracking module,in order to ensure that the algorithm can meet the real-time requirements on NVIDIA Jetson Nano,the YOLOv5n target detection algorithm is selected and improved.The results show that the improved YOLOv5n has a faster detection speed and requires only 2.7 GFLOPs on the premise of losing only 0.28%accuracy.For solve the problem of system false alarm caused by target surface sloshing and new bullet hole discrimination,the SORT target tracking algorithm is selected to track the bullet hole.The experimental results show that the algorithm can track the existing bullet in real time and identify the new bullet.(3)In the bullet hole ring value determination module,For solve the problem of misreporting the target caused by the perspective distortion in the input target image,the YOLOv5n algorithm is used to calculate the perspective distortion correction matrix to reduce the influence of perspective distortion on the target reporting accuracy.With the help of the recognition results of the ring value number and the loop point,the calculation of the target center and loop radius is completed,and the discrimination and detection results of the new bullet hole are used,combined with the extraction results of the effective region of the chest ring target to determine the effectiveness of the bullet hole.Finally,the ring value and offset o’clock direction of the effective bullet hole are obtained by using the bull’s-eye,the ring radius and the new bullet hole coordinates.3.In order to make the system more intelligent and convenient,this paper designs a handheld target reporting application and monitoring software on the shooting end and monitoring end respectively.Shooters can view the real target image,virtual target image and detailed shooting results in real time,and support to input shooters’ information to generate shooting results report with one click.Finally,this paper verified the target reporting performance of this system through 5 sets of simulation experiments.The results showed that the target reporting accuracy reached 98%.The system test was carried out using black and white breast ring target shooting target paper,which proved that the system has different color coatings.The adaptability of the target paper also proves that the overall performance of the designed target reporting software is good.The intelligent target reporting system designed in this paper has passed the actual use test.