基于小波多尺度积的目标识别

Target Recognition Based on Wavelet Multiscale Product

光学相关探测就是利用光学相关的方法,从混乱的图像中找出需要的目标,达到识别的目的。把小波变换应用于光电混合联合变换相关器,突破了传统的傅里叶变换的局限性,实现了对探测目标不同区域、不同尺度的分析。为了充分利用小波不同尺度的特性,采用小波多尺度积的方法提取出目标图像的边缘,兼顾了图像的细节与轮廓特征,将目标图像不同层次的轮廓信息和细节信息相结合,解决了复杂背景下目标图像的识别问题。光学实验结果表明,该方法有效增强了复杂背景目标的相关点强度,成功实现了目标的探测,具有良好的应用前景。

Optical correlation detection can discriminate target from clutter scene by the optical method to realize target recognition. Wavelet transform is applied to hybrid optoelectronic joint transform correlator （HOJTC）, it transcends the limitation of Fourier transform, and can analyze detected target in different quarters and scales. To exploit character of various scales, the method of wavelet multiscale product is proposed to extract edge feature of image. This method considers the detail and contour as well, and combines both of them at different levels. The problem of targets recognition with insufficient information is settled. Optical experimental results show that the method effectively enhances the energy of correlation peaks of target in clutter scene, successfully realizes target detection, and has a good prospect of application.