小波改进最大平均相关高度法实现畸变目标识别

Distorted target recognition based on wavelet-modified maximum average correlation height (MACH) algorithm

为了解决图像匹配识别中复杂背景下畸变目标无法准确识别的问题,结合墨西哥帽(Mexican Hat)小波母函数重新设计了最大平均相关高度(MACH)滤波器,得到墨西哥帽-最大平均相关高度(Mexican Hat-MACH)匹配滤波器进行复杂背景下畸变目标的识别。为了平衡墨西哥帽-最大平均相关高度匹配滤波器的各项性能,需要对原有的最大平均相关高度滤波器的3个性能控制参数和墨西哥帽小波函数的尺度系数重新进行选择和优化。通过对大量的不同目标进行实验后表明,当3个控制参数为γ=1,β=α=0.15,尺度系数为15时,新设计的墨西哥帽-最大平均相关高度匹配滤波器有较高的通用性,能够对各种类型的复杂背景下的畸变目标响应较高的相关峰,并有更高的畸变容差。作为实例,用墨西哥帽-最大平均相关高度滤波器对复杂背景下的小飞机目标和低对比度坦克目标进行了识别实验,验证了该方法的可行性。

In order to solve the problem of inaccurate recognition of distorted target under cluttered background in image matching recognition. Combined with the Mexican-hat mother wavelet, the Maximum Average Correlation Height （MACH） filter is re-designed. A Mexican hat- maximum average correlation height matched filter is obtained to recognize distorted target under cluttered background. In order to balance the performances of Mexican hat-maximum average correlation height matched filter, three performance control parameters of the existing MACH filter and the scale coefficient of Mexican bat wavelet need to be chosen and optimized again. Through plentiful experiments of different targets, results show that when the three performance control parameters are γ = 1 ,β = α= 0. 15 and the scale coefficient is 15, the new Mexican hat-maximum average correlation height matched filter has high versatility. It can generate higher correlation peak response and has higher distortion tolerance for various types of distorted targets under cluttered background. As experiment examples, a small plane and a low-contrast tank under cluttered background were recognized using the Mexican hat-maximum average correlation height matched filter. Experiment results prove the feasibility of this method.