图像阴影消除算法研究与实现.docx

摘要 本科毕业设计论文 题 目 图像阴影消除算法研究与实现 摘 要 由于阴影的存在，严重影响了运动目标的检测效果，对学者的研究造成了许多影响，所以对阴影进行检测并且消除已经成为运动目标检测中的关键问题之一。传统阴影消除方法对阴影消除有一些局限性，针对目标检测中的阴影问题，本文提出了一种新方法. 在本文中，提出了在HSV颜色空间下，对多个视频的阴影区域进行直方图统计分析，分别获得阴影在H、S、V通道下的各自的颜色特征，再根据以上得出的特征在相应的三个通道上使用阴影样本训练模型参数建立高斯阴影模型，在此基础上给出一种新的阴影消除算法。在前景检测的部分中，通过高斯混合模型得出前景图像，通过计算前景像素点与阴影模型的匹配程度对阴影进行判定和消除。 通过实验过程，与同类算法的对比分析表明：本算法对不同场景下的阴影消除是准确且实时的，与以往的阴影检测算法在阴影检测率和阴影区分率上均有提升。 关键词: HSV颜色空间 阴影特征 高斯模型 阴影消除  ABSTRACT 目录 Abstract The existence of the shadow seriously impacts the result of foreground object detection,it has a bad influence on our research, so detecting and eliminating the shadow is the key problem for the moving target detection. There are some limitations in the traditional shadow elimination methods,aiming to the shadow problem during the detection of the object,we think of a new histogram to statistic and analysis the color feature of the shadow under the HSV color space. In this paper, under the HSV color space, got the shadow feature of the H, S, V channel by histogram analysising the shadow area of a series of video. And then, we establish the Gaussian shadow model on each channel according to the statistic information, to obtain the model parameters through training the shadow sample. Finally, we propose a novel algorithm to eliminate the shadow based on the model we built . In the part of the foreground detection , we can get the foreground image by using the Gaussian mixture model , through computing how the pixel matched to the model to determine and eliminate the shadow. As a result of the experiment, comparing with the similar algorithms, the results show that the proposed algorithm can eliminate the shadow correctly in real-time under different scenarios, and it performs better on the metrics of shadow detection rate and the shadow discrimination rate compared to the existing algorithms. Key words: HSV color space shadow feature Gaussian model Shadow El