光学相关多目标识别技术

Multiple target recognition technology based on optical correlation

光电混合实时联合变换相关器能够实现目标的自动识别和定位(方位角和俯仰角),在复杂背景中,多目标动态测量一直是待解决的关键技术之一,由于背景复杂、目标多、噪声大,因而目标识别效果不佳。针对这一问题,提出了在光电混合联合变换相关器中应用高斯滤波和拉普拉斯算子相结合的新方法,并把其应用在目标图像的频谱面。即基于高斯滤波先对目标图像的频谱进行平滑、积分以滤掉噪声,再利用拉普拉斯算子进行边缘检测,使目标图像轮廓清晰,降低了噪声的影响,消除了双边缘现象,最大限度地保留了频谱图像的细节信息。该方法解决了多目标的识别问题,提高了光电联合变换相关器的目标识别率。给出了飞机多目标识别的实验结果。

Hybrid optoelectronic joint transform correlator （HOJTC） can realize real-time recognition and localization of objects （azimuth angle and altitude angle）. Multiple target dynamic measurement has been one of key technologies to be resolved in cluttered background. Sometimes even no correlation peaks appear because of cluttered scene, multiple target and large noise. Aimed at this problem, the combination of Ganssian filter and Laplace operator was used in the HOJTC for image enhancement of joint power spectrum. First, in order to reduce the noise, Gaussian filter was used to smooth and integrate the joint power spectrum, and then Laplace operator was used for edge detection. So that the target image contour could be clearer, the influence of noise could be reduced, and the phenomenon of double edge could be eliminated, which retained the details of spectrum image. This method solved the problem of multiple target recognition, and increased the rate of target recognition of HOJTC. The experiment results of two plane targets are shown.