基于蚁群优化与神经网络的坦克目标识别算法

Recognition algorithm of tank based on ant colony optimization and neural network

为了对成像引信探测得到的变形严重的图像进行识别,提出了基于蚁群优化与人工神经网络相结合的坦克目标识别算法.采用SUSAN特征检测原则提取目标图像的角点特征,作为神经网络模式分类器的输入.针对BP网络收敛速度慢,易于陷入局部极小点等问题,利用蚁群优化算法训练网络权值,可兼有ANN的广泛映射能力和蚁群算法的全局收敛以及启发式学习等特点.仿真实验表明,新算法能够有效缩短网络训练时间,提高目标识别精度.

In order to recognize the images detected by inferred imaging fuze,an algorithm based on ant colony optimization and neural network is proposed.The corners of the image are extracted by the SUSAN corner detection principle,and they form the input vector of the neural network.Since BP neural network has some problems such as slow convergence and easy trapping into the local minimum points,a combination of ant colony optimization with neural network is adopted.The method is characterized by its wide range mapping capabilities,global convergence and heuristic learning.The experimental results prove that the algorithm can shorten the training time effectively and increase the accuracy of recognition.