火箭喷管极性自动化测试系统的改进设计研究

使用计算机视觉方法进行的发动机极性自动化测试是火箭地面测试中重要的测试环节，该环节存在进一步改进和提升的空间。将递归全对场变换（Recurrent All-pairs Field Transforms, RAFT）光流算法替代传统光流法检测技术用于发动机喷管实时运动监测，并根据现场测试场景对光流算法进行了优化，提升了运动检测速度与测量精确度，使自动测试系统具备了摆角的估测能力；在软件系统设计层面，引入差异图像直方图法监听法辅助喷管动作识别，避免了光流法对于未处在监测流程中的摄像头的冗余监听资源消耗，降低了系统硬件设备的负载，同时实现了一种可视化在线判读软件的设计。提出的软件与算法方面的改进在当前已投入使用的极性自动化测试系统上实现了进一步的优化。

Research on Rocket Nozzle Polarity Automatic Identification System Improvement

The automatic test of engine polarity using computer vision method is an important test step in the rocket ground test, and there is room for further improvement in this step. In this paper, the RAFT deep optical flow algorithm is used to replace the traditional optical flow detection technology for real-time motion monitoring of the engine nozzle, and the optical flow algorithm is optimized according to the real test scene, which improves the motion detection speed and measurement accuracy, and makes automatic test system has the ability to estimate the swing angle; at the software system design level, this paper introduces the difference image histogram method to monitor the nozzle to assist the nozzle action recognition, which helps to avoid the redundant monitoring of the camera that is not in the monitoring process by the optical flow method resource consumption, and helps to reduce the load of system hardware equipment, and at the same time this paper realizes the design of a visual online interpretation software. The software and algorithm improvements proposed in this paper have been further optimized on the current polarity automated test system that has been put into use.