A real-time critical part detection for the blurred image of infrared reconnaissance balloon with boundary curvature feature analysis

Detecting and striking the critical part of blurred balloon image is an important approach to counterattack the reconnaissance balloons. The existing algorithms of the image target detection task are not able to achieve the high precision and real-time performance in the meanwhile since the critical part of the balloon is tiny, weak and not easy to be segmented. In this paper, a real-time algorithm based curvature feature in the polar coordinate system is proposed to detect the critical part of reconnaissance balloons. We divide the proposed method into three steps: the image is firstly subjected to a gray-scale projection by calculating third-order post-difference, then the balloon boundary is extracted in transformed polar coordinates, and finally the boundary curvature identifies the position of the critical part. The core strategy of the proposed method is to adopt the boundary features of the balloon instead of the general time-consuming image operations (e.g. region labeling, matching) to capture the target part. The experimental results show that the proposed method obtains high precision results with a real-time detection. Our proposed method achieves a processing speed of 200 frames per second on DSP (TMS320C6678) while a state-of-the-art detection precision (>93 $$\%$$ ), which overcomes the existing comparison algorithms.