Multi-mode Automatic Target-scoring System Design Based On Video Image Analysis

With the development of technology, the army is looking for a method to train soldiers in a more scientific way. Systems that can score shooting results automatically have been raised recently. Research in this area is developing in a very fast speed. This kind of system has following advantages:it keeps those result-readers out of danger, reduces soldiers’work-load and gets more accurate results. However, current auto-target-scoring systems cannot meet user’s demand perfectly. So, it will make a lot of sense to improve current system.This thesis proposes an auto-target-scoring system based on video analysis, which makes a lot of improvements and innovations. The system is equipped with a sensor to detect vibration. In terms of ordinary-shooting-mode, the system can get shooing results correctly in an adaptive way. The system has also made some optimizations in algorithm to work correctly under mobile-shooing-mode and night-shooing-mode. In the process of bullet-hole detection, this thesis proposes a method which combines differentiating method and template-matching method. This method helps to reduce noise remarkably. What’s more, this thesis uses optical flow method and bullet shape feature calculation to ensure that every new bullet-hole is detected. Lastly, an online-learning method is used, which can update detecting parameters after every shoot. As a result, the system gets more and more accurate.The auto-target-scoring system uses system-on-chip (SOC) which is designed by TI. The SOC has great performance on digital media applications, and we can design software easily thanks to its development toolkit. The DSP on SOC, which is called C674X, is used to run the entire algorithm. M3VPSS is used to image signal capturing, while M3VIDE0 is used to image encoding. Cortex-A8 is used to control the whole system, to communicate with other parts of the system.According to the test, the system has a great performance under each shooing mode. The system can be applied into real use under ordinary-shooing-mode, and also works well in simulation environment under the other two modes. In conclusion, system proposed in this thesis can satisfy real shooting-training requirements.