基于改进动态规划的主动声纳序列图像检测方法

动态规划是序列图像小目标检测的常用方法，该方法假设各帧图像间目标转移过程为马尔可夫过程，将能量累积计算量由原本的按图像帧数幂指数增长变为成倍增长，大大减小了计算量，使计算目标多帧累积能量变为可能。为了提高算法检测能力，该文根据动态规划算法原理及惩罚函数特点对算法进行了改进，使原本的二阶规划过程变为三阶，使指标函数更加有效地累积，有效提高了相同虚警概率下的检测概率。还利用该算法设计了主动声纳序列图像检测过程，并进行了仿真，验证了算法的有效性。

An active sonar sequence image detection method based on improved dynamic programming

Dynamic programming is a common method of small target detection by sequence images. It assumes that the transfer processes between frames are markov process to make the calculation amount of energy accumulation process turn from exponent growth into multiply growth by the number of frame, and greatly reduce the calculation amount, which make it possible to calculate the accumulated energy. To improve the detectability of the algorithm they converting the original two stages planning process to three stages according to the dynamic programming algorithm principle and characteristics of penalty function, so that the indicator function can be accumulated more effectively and improve the detection probability under different false alarm probability. In addition, the detection algorithm by active sonar sequence images based on the improved dynamic programming is designed. And the effectiveness of the algorithm is verified by the simulation.