论文部分内容阅读
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
When using the optical pattern recognition technology and SDF filtering technique to detect the 3D target distortion in the actual scene, due to the recognition of a large number of non-training images and the difference in scene images Noise, background interference and various non-ideal features of the hardware identification system inevitably lead to serious degradation of the S / N of the relevant planes, so that the method of dividing the relevant signals by the usual threshold technique fails. Thus greatly reducing the recognition efficiency of the OPR system. This article uses artificial neural network (ANN) technology and optical pattern recognition technology (OPR) combination. By partitioning the region of interest (ROI) and extracting the intensity distribution of the relevant plane and the training process of the offline artificial neural network, the OPR system can effectively give the input training images, the non-training images and various background noises respectively Different output response, so as to achieve a greatly improved OPR system, the correct recognition rate