针对水下场景目标探测图像质量退化问题, 提出了一种自适应计算水体衰减系数暗通道融合多尺度Retinex(Multi-scale Retinex, MSR)的复原算法, 有效实现了水下目标的复原。通过搭建的水下成像测量装置, 借助成像系统获取水下模拟环境的探测图像, 对水下探测图像按照算法流程图逐步处理, 得到了有效复原水下目标辐射信息的图像。为客观评价算法的效果, 采用对比度、平均梯度与信息熵作为定量评价指标因子, 对该算法与常规三种算法进行了定量对比研究, 结果表明, 该算法处理结果各项定量评价指标因子均优于选取的对比算法。研究结果为水下目标探测提供了基础理论探索方法, 对水下目标探测实施开展具有一定的指导意义。

For the problems of image quality degradation in underwater scene target detection, an algorithm which combined the improved dark channel with MSR was proposed, which could adaptively compute water attenuation coefficient and effectively realize the recovery of underwater target. Through the built-in underwater imaging measurement device, the detection image of the underwater simulated environment with the aid of imaging system was obtained, the underwater detection image was processed step by step according to the algorithm flow chart, and an image for the effective recovery of underwater target radiation information was obtained. In order to objectively evaluate the algorithm effect, contrast, average gradient and information entropy were adopted as quantification to evaluate indexes factors. A quantitative comparison study between this algorithm and the conventional three algorithms was performed. The result show that the improved algorithm to deal with the results is better than the selected compared algorithms under all the quantitative evaluation indexes factors. The research results provide a basic theoretical exploration method for the underwater target detection, as well as have certain guiding significance for the implementation of underwater target detection.