【摘要】 在采用光学模式识别技术、ＳＤＦ（综合鉴别函数）滤波技术进行实际场景中的三维目标畸变不变识别的时候，由于面对的是大量的非训练像的相关识别，加上场景图像中的不同噪声、背景的干扰，以及硬件识别系统的各种非理想特征等因素，均不可避免带来相关平面的Ｓ／Ｎ的严重退化，从而使按通常的阈值技术进行相关信号分割的方法失败。因而大大降低了ＯＰＲ系统的识别效率。本文采用人工神经网络（ＡＮＮ）技术与光学模式识别技术（ＯＰＲ）相结合。通过对相关平面感兴趣区域（ＲＯＩ）的分割与强度分布特征抽取以及脱机人工神经网络的训练过程，使ＯＰＲ系统能有效地对输入的训练像、非训练像及各种背景噪声分别给出不同的输出响应，从而达到了大大提高ＯＰＲ系统的正确识别率更多还原

【Abstract】 When making the 3-D object distortion invariant correlation recognition using Optical Pattem Recognition(OPR) and Synthetic Discriminant Function(SDF) filtering technique,the correlation S/N will be degraded severely because of the three factors:nontraing images correlation,interference of background and clutter,and nonideal characteristics of OPR hardware system.As result,the failure of correlation signal extraction by thresholding technique occurs,the recognition possibility of the OPR system will be decreased greatly. A Artificial Neural Network (ANN) technique was used to improve the recognition possibility of the OPR system,to which the different output was got to the input of training,non training image and various background and clutter.Through the ROI extraction,contour mapping and off line training procedure of ANN,the correlation signal classification was performed efficiently.更多还原