低能见度条件下的交通图像增强技术研究

发布时间：2018-03-13 09:27

本文选题：低能见度　切入点：小波变换　出处：《天津职业技术师范大学》2016年硕士论文　论文类型：学位论文

【摘要】：随着信息技术的快速发展,智能交通成为交通运输管理发展的必然。交通监控系统是智能交通系统的一个重要手段,已经具有网络化、智能化、系统化和自动化等优点,这使得交通管理人员不必亲临现场或对交通录像进行逐帧分析,就可以直接找出违规车辆现场图像。但在实际应用中,交通视频监控系统大部分是以晴天等正常能见度为前提条件设置的,受雾霾天气、夜间等低能见度条件影响,监控系统难以采集到清晰高质量的视频图像,这种情况下采集的图像难以进行有效处理。为提高低能见度图像的对比度,使采集到的图像能够有效对车辆、道路标志标线的识别,本文对雾霾和夜间两种低能见度条件下采集的图像进行了图像增强研究,分别提出了雾霾图像增强方法和夜间图像增强方法,其主要内容如下:1)根据雾霾交通图像的特点,本文提出了一种基于小波变换和Retinex双边滤波的图像增强算法。首先对HSV模式图像V图层进行小波分解,然后对低频部分进行系数增强和基于双边滤波的Retinex算法处理,对高频部分进行阈值去噪,这样不仅有利于降低算法复杂度,而且可以在去噪的同时较好的保持图像的细节。双边滤波在处理相邻像素值时,同时考虑像素的邻近关系和亮度相似性,能够保持图像边缘不被模糊、抑制光晕效应的产生。再然后,对重构后的V图层进行全局对比度增强,进一步提升了图像的对比度。最后,饱和度增强使图像颜色更加鲜艳,增强效果更好。通过实验比较,本文方法与其它增强算法相比,图像在整体亮度和对比度方面明显改善,细节更加丰富,边缘亮度也有一定的增强,针对交通图像增强具有非常好的效果。2)夜间交通图像相对于雾霾交通图像更加复杂,尤其是缺少光照的交通路口图像,由于光照严重不足,图像中存在大量的暗连通区域,为此本文提出了一种针对夜间图像的增强方法。与雾霾图像增强算法不同,对V图层进行小波分解前,先进行全局对比度增强的预处理,以提升图像的整体亮度,这样也使夜间图像的反转图像与雾霾图像特征类似,可以采用雾霾图像增强的方法进行处理。对小波分解后的低频部分及低频部分的反转图像分别进行自适应权值的MSR算法处理。自适应权值的MSR算法能更好地调节亮度。为了在图像中更好地显现路面等固定场景,采用白天相同场景下的视频图像提取图像背景,经小波分解后的低频部分与夜间图像小波分解后低频部分加权,这样不仅能使图像显现固定场景,而且由于只对低频部分进行操作,对夜间图像细节影响较小。实验证明,本文算法能够有效地提升夜间图像的亮度、对比度,使图像色彩自然,并且能够避免光晕现象的产生。
[Abstract]:With the rapid development of information technology, intelligent transportation has become the inevitable development of transportation management. Traffic monitoring system is an important means of intelligent transportation system, which has the advantages of network, intelligence, systematization and automation. This makes it possible for traffic managers to directly find out the scene images of illegal vehicles without having to visit the scene in person or to analyze traffic videos in frame by frame. But in practical applications, Most of the traffic video surveillance systems are set on the premise of normal visibility such as sunny days. Affected by the low visibility conditions such as haze weather and night, it is difficult for the surveillance system to collect clear and high-quality video images. In order to improve the contrast of the low-visibility images, the collected images can effectively identify the vehicle and road marking lines. In this paper, we study the image enhancement of haze and nocturnal images under low visibility, and propose haze image enhancement method and night image enhancement method, respectively. The main contents are as follows: 1) according to the characteristics of haze traffic image, In this paper, an image enhancement algorithm based on wavelet transform and Retinex bilateral filtering is proposed. Firstly, the V layer of HSV mode image is decomposed by wavelet transform, then the low frequency part is enhanced by coefficients and the Retinex algorithm based on bilateral filtering is processed. The threshold denoising of the high frequency part is not only helpful to reduce the complexity of the algorithm, but also to keep the details of the image at the same time. Considering the pixel proximity and brightness similarity, the edge of the image can be kept unblurred, and the halo effect can be suppressed. Then, the global contrast of the reconstructed V layer is enhanced. Finally, saturation enhancement makes the image color brighter and the enhancement effect better. Compared with other enhancement algorithms, the method in this paper improves the overall brightness and contrast of the image obviously. The details are more abundant, the edge brightness is also enhanced, the traffic image enhancement has a very good effect. 2) the night traffic image is more complex than the haze traffic image, especially the traffic intersection image with lack of light. There are a lot of dark connected regions in the image because of the serious lack of illumination. In this paper, an enhancement method for night image is proposed. Different from the haze image enhancement algorithm, before the wavelet decomposition of V layer, In order to improve the overall brightness of the image, the inversion image of the night image is similar to that of the haze image, so that the whole image is enhanced by the preprocessing of the global contrast enhancement. We can use the method of haze image enhancement to process the low frequency image and the inversion image of the low frequency part after wavelet decomposition. The MSR algorithm of adaptive weight can be used to adjust the low frequency part and the inversion image of low frequency part. The MSR algorithm of adaptive weight can be adjusted better. Brightness. In order to better display the road surface and other fixed scenes in the image, The video image of the same scene during the day is used to extract the background of the image. The low-frequency part of the image is weighted by wavelet decomposition and the low-frequency part of the night image is decomposed by wavelet, which can not only make the image appear fixed scene. The experiments show that the algorithm can effectively improve the brightness and contrast of the night image, make the image color natural, and avoid halo phenomenon.
【学位授予单位】：天津职业技术师范大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41

【相似文献】