基于视频分析的停车场车位检测实验系统

发布时间：2018-06-07 08:21

本文选题：车位检测 + 灰度直方图　；参考：《海南大学》2016年硕士论文

【摘要】：目前,在经济的推动下,国民生活水平逐渐走向富裕,汽车的数量也在逐年增长,停车难的问题日益严峻。如何通过现代科技,使停车位资源得到充分利用,已经成为一个研究热点。本文对智能停车场的国内外研究现状做了简要分析,并对目前比较常见的一些车位检测方法做了对比。与非视频车位检测技术相比,利用视频图像进行车位检测具有明显的优越性。本文针对传统的车位检测算法的弊端,提出一种基于视频图像处理的检测算法。这种方法整合车位图像的灰度特征和纹理特征,将之作为判断车位状态的依据。特征的处理方式,是车位检测技术的关键。提取各个车位图像的灰度特征和纹理特征,将之组合成联合特征向量。采用主成分分析法对车位的联合特征向量进行降维处理,并采用支持向量机对降维后的特征向量进行训练。使用训练得到的模型对需要进行车位检测的视频图像分类,从而得到车位是否被占用的信息,并将得到的车位信息实时显示。为了使支持向量机训练出更符合实际情况的模型,在训练样本中加入了含有不同天气、光照、行人、杂物及阴影的视频图像。经过验证,该算法检测准确率为98.150%,能够排除天气、光照、车位上的行人、杂物、阴影等对检测结果的影响。且该算法能同时处理多个车位,大大提高了停车场车位管理的效率。最后,本文利用MATLAB对此车位检测方法进行了模拟仿真,并在仿真的基础上构建了车位检测实验系统。
[Abstract]:At present, driven by the economy, the national standard of living is gradually becoming rich, the number of cars is also increasing year by year, and the problem of parking difficulty is becoming more and more serious. How to make full use of parking space resources through modern science and technology has become a research hotspot. This paper makes a brief analysis of the domestic and international research status of intelligent parking, and compares some common methods of parking space detection. Compared with the non-video parking detection technology, the use of video images to detect parking space has obvious advantages. In this paper, a detection algorithm based on video image processing is proposed to overcome the disadvantages of the traditional parking space detection algorithm. This method integrates gray and texture features of parking space image as the basis for judging parking space state. The processing method of feature is the key of parking spot detection technology. The grayscale feature and texture feature of each parking space image are extracted and combined into a joint feature vector. The principal component analysis (PCA) is used to reduce the dimension of the joint eigenvector and the support vector machine (SVM) is used to train the dimensionally reduced eigenvector. The training model is used to classify the video images that need to be detected to get the information of whether the parking space is occupied or not and display the parking space information in real time. In order to train support vector machine (SVM) to fit the actual situation, video images with different weather, light, pedestrian, clutter and shadow were added to the training samples. The accuracy of the algorithm is 98.150, which can eliminate the influence of weather, illumination, pedestrians, sundries and shadows on the detection results. The algorithm can deal with multiple parking spaces at the same time, which greatly improves the efficiency of parking space management. Finally, this paper simulates the method of parking spot detection by using MATLAB, and builds an experimental system of parking space detection on the basis of simulation.
【学位授予单位】：海南大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41

【参考文献】