高斯模型与区域生长相结合的景物识别算法  

Natural Object Recognition Algorithm Using the Combination of Gaussian Model and Region Growing

作　　者：雷宝权[1] 程咏梅[1] 杨丽华[1] 赵春晖[1] 

机构地区：[1]西北工业大学自动化学院,陕西西安710072

出　　处：《计算机仿真》2010年第11期262-266,共5页Computer Simulation

摘　　要：研究景物图像特征提取,光照环境因素变化时,室外场景光学成像也随之发生很大变化,景物特征复杂,严重影响了景物识别的精度。为了解决上述问题,提出一种高斯模型与区域生长相结合的景物识别算法。先提取图像区域的底层视觉特征以及空间位置特征,并通过高斯分布为每类景物建立模型,然后根据模型测试图像中区域属于每一类景物的概率,把概率值较大的区域加入到模型中,更新模型参数,最后,把概率值较大的区域作为种子点进行区域生长得到景物识别的结果。采用Matlab 7.0软件对Pasadena数据库中景物进行仿真识别。结果表明识别目标景物的有效性,为复杂场景下景物识别提供了一种有效的算法。As outdoor scene images change greatly with the light and other environmental factors,the accuracy of the natural object recognition algorithm is affected seriously.In order to settle this problem,an algorithm using the combination of Gaussian model and region growing is presented.Firstly,visual features of the training images are modeled through Gaussian model,and the prior information about spatial location is joined as well.Then,the probabilities of regions belonging to each scene are tested,the regions with large probability values are added to the model,and the model parameters are updated at the same time.Finally,the region growing algorithm is used to obtain the recognition results.The algorithm is tested on the images from Pasadena Houses2000 database including 5 categories of natural objects such as sky,road,house,tree and grass,and satisfying results are achieved.Experimental results demonstrate the superiority and a better identifying accuracy of the algorithm proposed.This research provides an effective algorithm for natural object recognition in complicated scene images.

关 键 词：景物识别 高斯模型 区域生长 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]