基于边缘区域不变矩的缺损扩展目标识别方法

Occluded extended target recognition using moment invariants based on edge region

该方法提出以基于边缘区域的局部不变矩作为识别特征,结合多神经网络实现对缺损扩展目标的有效识别。讨论了离散情况下基于边缘区域局部不变矩的平移、旋转和尺度不变性。在此基础上,建立目标多个处理区域的BP人工神经网络,利用各网络分类综合结果提高缺损目标的识别率。实验结果显示该方法能够对缺损扩展目标进行正确识别,特别对于有较大部分缺损的扩展目标识别有明显优势。

Using the local moment invariants based on edge region as the recognition feature and building multiple neural networks, a novel method for partially occluded extended target recognition is presented. First, the local moment invariants are calculated on partial edge region of a target and their invariance in digital condition is discussed. Then, multiple BP neural networks are built on one or several local areas of the occluded extended target, and the moment invariants based on edge region of these local areas are calculated as inputs of neural networks. Through the integrated result of multiple neural networks, the correct recognition ratio can be improved. The experimental results indicated that the moment invariants based on edge region are simple and valid, and the proposed method can recognize partially occluded extended targets correctly, Especially for targets with large occluded part.