Real-time preprocessing for vision measurement system on high-speed cooperative target

In the vision measurement of high-speed moving target, the system should with high data-update rate and maintain it stable. Cooperative markers also should be correctly matched. As the markers on the moving target with large displacement in the images between frames, the conventional neighborhood search method with a fixed neighborhood size would decrease the system data-update rate, mismatch the markers, and even miss the target, which result from inappropriate neighborhood setting. In this paper, an embedded real-time preprocessing system is designed. The problem of low data-update rate caused by high-resolution images transmission has been solved by Zero-Delay data transmission mode. A self-adaptive neighborhood search method is proposed with Kalman Filter to estimate the position of neighborhood, and the feedback of the depth information in the measurement results is used to track the neighborhood size. Benefiting from this method, the search range of the marker between frames is remarkably reduced. In the experiment, at the distance of 1.9m, the data update rate of the system remains relatively stable above 65fps when the moving target velocity ranged from 0.033m/s to 2.01m/s. The advantages in matching and real-time ability of the system in high speed target measurement are verified.