基于距离选通成像的水下线状目标检测算法

摘要

An algorithm for long distance underwater linear object detection based on range-gated laser imaging is proposed.For the characteristics of underwater images,such as low contrast,blur and noise,and so on,the image was enhanced by means of contrast stretching,median filtering and wavelet transform.Then,the Canny operator was used to extract object's edge features.Finally,for the emergence of noise edges,a robust algorithm named Random Sample Consensus was chosen to detect linear object and estimate its parameters such as position and direction.The experiment results show that the algorithm can effectively detect underwater linear objects,and the detection rate can reach 93％,and the effective detection range can be up to 5 times the length of the underwater decay.%基于水下距离选通激光成像技术,提出了一种可用于长距离下的水下线状目标检测算法.该算法针对水下成像中低对比度、模糊和噪声等特性,首先采用对比度拉升、中值滤波、小波变换等方法对图像进行增强处理;然后利用Canny边缘检测算子提取出目标的边缘特征;最后针对边缘特征中出现的噪声边缘问题,选用了鲁棒性强的随机抽样一致性参数估计算法从边缘特征中检测出线状目标,并计算得到目标的位置和方向等相关参数.实验结果表明,该算法可以有效地检测出水下曲线状目标,弥补现有方法只能检测直线目标的不足,检测率可以达到93％,有效检测距离能达到5倍水下衰减长度.