针对间断纹理环境中的图像特征追踪和匹配算法研究

针对同时定位与地图构建(SLAM)中的特征匹配关键环节,提出一种融合特征点和特征区域的图像追踪与匹配算法,以解决交替出现纹理丰富和纹理缺失的间断纹理环境中图像特征易丢失、误匹配率较高的问题。首先,利用ORB算法和半稠密直接法分别对图像提取特征点和特征区域。其次,使用渐进一致采样法(PROSAC)剔除ORB算法的误匹配特征点,并计算特征点的正确匹配率。最后,针对纹理缺失环境中特征点丢失严重的问题,以特征点的正确匹配率作为判断依据,对低匹配率图像,则基于特征区域使用半稠密直接法实现图像的追踪,同时对追踪结果进行非线性优化,提高了特征区域追踪的准确性和稳定性。实验结果表明,该算法适用于间断纹理环境,在纹理丰富和纹理缺失条件下均可提高图像匹配的准确率。

Research on Image Feature Tracking and Matching Algorithms in Intermittent Texture Environment

Aiming at the key step of feature matching in SLAM, an image tracking and matching algorithm combining feature points and feature regions is proposed to solve the problem of easily losing image features and high mismatching rate in discontinuous texture environment where texture is rich and texture is missing alternately. Firstly, ORB algorithm and semi-dense direct method are used to extract feature points and feature regions respectively. Secondly, progressive uniform sampling (PROSAC) is used to eliminate the mismatched feature points of ORB algorithm, and the correct matching rate of feature points will be calculated. Finally, aiming at the serious problem of feature point loss in texture-missing environment, the correct matching rate of the feature point should be the judgement basis. As for the low-matching images, a semi-dense direct method should be used to track low-matching image based on the correct matching rate of feature points. At the same time, the tracking results are optimized nonlinearly to improve the accuracy and stability of feature area tracking. The experimental results show that the algorithm is suitable for discontinuous texture environment, and it can improve the accuracy of image matching on the conditions of rich texture and texture missing.