基于蛙人探测声纳序列图像的水下小目标检测算法  

Underwater Small Target Detection Algorithm Based on Diver Detection Sonar Image Sequences

作　　者：毛盾[1] 刘忠[1] 程远国[1] 

机构地区：[1]海军工程大学电子工程学院,武汉430033

出　　处：《传感技术学报》2011年第7期1027-1032,共6页Chinese Journal of Sensors and Actuators

基　　金：中国高科技研究发展计划(863)项目(2007AA01Z309);中国博士后科学基金项目(20090461460)

摘　　要：针对水下监控系统的小目标检测问题,在分析蛙人探测声纳成像特点的基础上,提出了基于图像二值化和区域生长法的自适应双帧差法。通过将双帧差结果与当前帧的二值化图像进行"与"运算来消除"双影",然后采用区域生长法消除"空洞",最后根据检测结果自适应地调整阈值。克服了三帧法和累积帧差法在处理空洞和双影问题时存在较长时延的缺陷,提高了水下低信噪比环境下对小目标的检测能力。实验证明了该算法的有效性。To automatically detect small targets in diver detection sonar,a self-adaptive inter-frame differencing algorithm based on image binarization and region growing was presented.It removed firstly the ＂bi-image＂ in differencing images using the binary image of current frame,and then filled up the target ＂hole＂ in differencing images through region growing method.Finally,the algorithm parameters were adjusted according to the detection result.Experiments show it overcomes Three-Frame-Differencing and Accumulative-Frame-Differencing methods＇ weakness that they waste too much time in resolving the problems of bi-image and hole,improves the ability of small target detection under underwater low SNR environment,and is a practical method.

关 键 词：自动目标检测 蛙人探测声纳 差分图像法 区域生长法 自适应阈值 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术] TB566[自动化与计算机技术—计算机科学与技术]