为了抑制ViBe算法在海面动态背景视频下“鬼影”区域对船舶运动目标检测的影响,提高监控视频中船舶运动目标识别的准确率,提出一种改进的ViBe算法。首先,背景模型用连续帧初始化,以减少“鬼影”的影响;然后,使用自适应阈值和闪烁级别来减少海面杂波,同时采用像素点对比消除“鬼影”,提取运动目标前景,获取完整的运动目标区域。最后,对输入视频进行高斯金字塔多尺度分解,并采用改进的ViBe算法检测低分辨率视频中的移动船舶,完整提取了海上移动船舶。实验结果表明,所改进的算法消除了“鬼影”区域,减少了海面杂波的干扰,检测率为92.5%,单帧视频图像检测时间控制在97 ms以内,可准确、快速地检测和提取海面船舶运动目标。
In order to suppress the influence of ghost area to the ViBe algorithm on the detection of ship moving objects under the dynamic background video of the sea surface and improve the accuracy of the ship moving object recognition in surveillance video, an improved ViBe algorithm was proposed. First, the background model is initialized with continuous frames to reduce the influence of ghost. Then, the sea clutter is reduced by using adaptive thresholds and flicker levels. Third, the ghost elimination strategy is used to eliminate ghost. Finally, Gaussian pyramid multi-scale decomposition is performed on the input video, and the improved ViBe algorithm is used to detect moving ships in low-resolution video, which further reduces the interference of sea clutter and completely extracts moving ships. The detection time of a single frame of video image is controlled within 97 ms, which can meet the requirements of real-time detection.
2023,45(5): 164-169 收稿日期:2022-03-10
DOI:10.3404/j.issn.1672-7649.2023.05.032
分类号:U675.7
基金项目:国家自然科学基金资助项目(71774019);中央高校基本科研专项资金(017210127)
作者简介:杨家轩(1981-),男,博士,副教授,研究方向为交通信息工程及控制、航海安全保障