红外航空图像自动目标识别的形态滤波神经网络算法

 提出了一种有实用意义的形态滤波神经网络模型及其自适应 BP学习算法。形态滤波网络的优化设计过程实际上是网络参数 (结构元素 )不断调整、逐步适应图像环境的优化学习过程,从而将目标客体的特征规律反映到网络结构上来,以实现对复杂变化的图像具有良好的滤波性能和稳健的适应能力。为结合运动图像目标的检测需要,采用了渐进收缩误差、适时校正网络权值的动态跟踪学习算法。通过实验结果可以看出,该算法不仅能适应复杂多样的背景环境,而且对运动目标的连续检测能力具有位移不变、伸缩不变和旋转不变的特性。

MORPHOLOGICAL NEURAL NETWORKS WITH APPLICATIONS TO AUTOMATIC TARGET RECOGNITION IN AERONAUTICS INFRARED IMAGE

A practical neural network model of morphological filters with its optimal parameters training algorithm is proposed. For application to the infrared motional image target detection, a dynamic training algorithm, namely Morphological Adjusted\|Weight Neural Network (MANN), is applied to the detecting process using the asymptotic shrinking error and appropriate network weights adjusting. Experimental results show that the algorithm has an invariant property with respect to the shift, scale and rotation of the moving target in the continuing detection of moving targets.