关键词: 港口航道工程；船舶号灯识别；目标检测；背景差分法；动态场景；SURF算法

Abstract: Accurate detection and identification of ship signal lights is one of the effective ways to achieve situational awareness of marine vessels. A detection and identification method of ship signal lights in dynamic scenes based on background motion compensation and optimized background subtraction method was proposed. Firstly, based on the SURF feature point extraction algorithm, circular regions were used to replace rectangular regions to extract 32 dimensional descriptors, achieving dimensionality reduction of descriptors and improving the speed of the algorithm. Secondly, the refined SURF algorithm was utilized to extract and match feature points from video images, yielding linear parameters that reflected the mapping relationship between images. These parameters were then used for background estimation and completing background motion compensation. Finally, the background subtraction method was optimized by adopting a segmented updating strategy and adaptive difference threshold, and the geometry and color characteristics of the signal lights were combined to eliminate the influence of environmental factors such as interference lights and sea waves. The research results show that the proposed algorithm after completing background motion compensation has not only high detection and identification accuracy of ship signal lights but also strong robustness, which can better detect and identify ship signal lights in dynamic backgrounds.

Key words: port and waterway engineering; ship signal lights identification; object detection; background subtraction method; dynamic scenes; SURF algorithm