室外场景下目标分割和目标识别算法

视频监控应用场听可分为室内和室外．室外环境受到诸如光照、下雨、落叶等因素的影响，整个场景变化复杂，给视频处理带来许多困难．文中介绍一种室外场景下目标分割和目标识别的方法，使用基于像素颜色特征的混合概率模型，将当前图像中与模型匹配的像素视为背景，然后更新模型中各个参数．为了去除目标区域的阴影部分，引人一种基于阀值和区域特征的阴影消除算法．同时，采用基于支持向量机的分类方法，识别场景中新出现的目标．

An Algorithm for Object Segmentation and Object Recognition on Outdoor Scene

The site where video monitoring is applied can be divided as indoor and outdoor. The outdoor surroundings bring about difficulties to video-frequency processing due to the influence of such factors as illumination, raining and fallen leaves which make the entire scene to change and to be complicated. A method is presented here for object segmentation and object recognition. The author regards the pixels in present image matching with a model of mixed probability based on the characteristic of pixel color as background, and updates parameters in the model; and eliminates the shadow in object area by leading in an algorithm based on threshold and characteristic of area; and moreover, recognizes new-emerging object in the scene by adopting method of classification based on a supporting vector machine.