【作者】 苗华；

【导师】 王文生；

【作者基本信息】 长春理工大学 ， 光学工程， 2009， 博士

【摘要】 光学图像相关探测和识别,是对采样点并行处理,并以光速实现图像信号的傅立叶变换,通过计算机对图像信号进行数字处理。论文把现代光学测试技术与现代探测器件相结合,研制了光电混合联合变换相关器,实现了光学图像的探测、实时跟踪、自动识别和高精度定位。然而,实际的目标图像背景非常复杂,对比度较低,亮度不足,常常导致探测和识别能力降低,甚至无法识别。因此对物面图像和谱面进行数字图像处理,提高低对比度目标的识别率,是关键技术之一。为增强相关峰对比度,在谱面采取了墨西哥帽小波变换,此方法有效地抑制了联合变换功率谱中的噪声信息,提高了有用信息,增强了大部分相关峰对比度。针对低对比度目标识别问题,在物面对于静态目标分别采取了对比度拉伸法、高斯小波模极大值边缘检测法、高斯小波边缘检测与数学形态学结合处理方法、数学形态学与阈值结合处理方法、改进的Robert算子边缘检测处理方法,有效地抑制了物面图像中的背景噪声干扰,增强了绝大部分相关峰对比度,提高了复杂背景低对比度静态目标相关识别能力。实验表明对于水上低对比度目标高斯小波边缘检测与数学形态学结合处理方法获得的相关峰图像中相关点的能量相对其他几种处理方法来说增强最明显,并且背景噪声最弱,其获得的相关峰对比度最好。对于陆地低对比度目标高斯小波边缘检测与数学形态学结合处理方法和改进的Robert算子边缘检测处理方法获得的相关峰图像中相关点的能量相对其他几种处理方法来说增强最明显,并且背景噪声最弱,其获得的相关峰对比度最好。对于动态目标研究了物面自动阈值处理方法,经自动阈值处理后有效地滤除了物面图像中的背景噪声,增强了相关峰对比度,实现了光电混合联合变换相关器对复杂背景低对比度运动目标的自动探测识别。更多还原

【Abstract】 Optical correlation detection and recognition is to process all sampling points at the same time and realize Fourier transform of image signals with light speed. Modern optical test technology is combined with modern detecting devices in this paper. Optoelectronic hybrid joint transform correlator is developed successfully and detection, real-time tracking, automatic recognition are realized with high location accuracy. However, real targets have relative complex background, low contrast and insufficient brightness, which usually lead to decrease the ability of target detection and recognition and even miss the target. Therefore, one of the key technologies is to implement image processing of object image and power spectrum image to improve recognition rate of targets.To enhance the contrast of correlation peaks, Mexico-hat wavelet transform is adopted in power spectrum plane, with which noise information of joint transform power spectrum is suppressed successfully. Useful information is increased and the contrast of correlation peaks is also strengthened.As to the static targets recognition, many methods are adopted in the object plane to suppress the noise interference of background effectively, improve contrast of most correlation peaks and strengthen correlation recognition ability of static targets in cluttered background with low contrast. The methods include contrast stretching, edge detection by Gauss wavelet modulus maximum value, combination of edge detection by Gauss wavelet with mathematical morphology, combination of mathematical morphology with threshold processing and edge detection by improved Robert operator.As to targets on water with low contrast, experiments show the energy of correlation peaks are stronger with the combination method of mathematical morphology and edge detection by Gauss than the other methods. Correspondingly, background noise is weaker and contrast of correlation peaks is also better.As to targets on land with low contrast, experiments show the energy of correlation peaks are stronger with the method of edge detection by improved Robert operator and combination method of mathematical morphology and edge detection by Gauss wavelet. Thus, background noise is weaker and contrast of correlation peaks is also better.As to dynamic targets, the processing method of automatic threshold in object plane is researched, with which background noise is filtered out effectively in object plane and contrast of correlation peaks is also strengthened. Therefore, detection and recognition of moving targets with cluttered background and low contrast is realized successfully by optoelectronid hybrid joint transform correlaor.更多还原