基于Markov随机场的三维物体识别算法

Recognition of multiple 3-D objects based on Markov random field models

为准确识别出三维物体,提出了一种新的物体特征框架,采用密集采样的多分辨率网格来描述物体观测图像的局部特征,引入Markov随机场模型对网格节点之间的几何关系进行建模。不同图像之间的匹配通过最高置信度优先算法实现,以获取两图像各个节点之间的准确匹配关系以及全局相似度。在Coil-100(columbiaobjectimagelibrary)图像数据库上,以100个物体的4、8、18、36个视角的样本为模板,用其他68、64、54和36个视角的样本进行测试,该算法识别率分别为95.75%、99.30%、100.0%和100.0%,识别准确率明显高于文献中的方法,这说明算法在基于观测图像的物体识别领域有着非常好的应用前景。

Computer vision systems can not easily identify 3-D objects. This paper presents an object framework which utilizes densely sampled grids with different resolutions to represent the local information of the input image. A Markov random field model is used to model the geometric distribution of the key object nodes. Flexible matching, which seeks to find an accurate correspondence map between the key points of two images, combines the local similarities and the geometric relations using the highest confidence first method. Then, a global similarity value is calculated for the object recognition. The algorithm was evaluated using the Coil-100 object database, which consists of 7 200 images of 100 objects. When the numbers of templates for each object were varied from 4, 8, 18 to 36, the object recognition rates were 95.75%, 99.30%, 100.0% and 100.0%, which are much higher than those of previous algorithms. This excellent recognition performance indicates that the approach is well-suited for appearance-based object recognition.