基于LOG GABOR小波相位一致不变量的目标识别

Target Recognition Based on Log Gabor Wavelet Phase Congruency Feature Invariant

给出了一种基于LOG GABOR小波的相位一致(PC)不变量的神经网络目标识别方法。针对GABOR小波存在的问题,分析了LOG GABOR小波优于GABOR小波的性能,给出了PC不变量的定义,探讨了低层次图像不变量特征,运用LOG GABOR小波PC特征不变量公式进行了修正,提取了目标图像边缘特征。利用该方法进行了神经网络目标识别实验,仿真结果表明,该方法能够很好识别图像目标,识别率达到97%。

The method of neural network target recognition was presented based on Log Gabor wavelet phase congruency（PC）. The performance of Log Gabor wavelet was excelled to that of Gabor wavelet in theory. The definition of PC invariant was presented,and the low-level image invariant feature was discussed. The formula of PC feature invariant was modified by Log Gabor wavelet. The experiment of image target recognition was done by the PC. The simulated result indicated the method based on the PC feature invariant of Log Gabor wavelet can finely recognize the image target,and the recognition rate attained to 97%.