亮度变化下室外场景图像特征点提取方法

Method of Extracting Feature Points from Outdoor Scene Images with Variable Brightness

在室外场景的环境感知中,固定阈值的特征点提取算法,其特征点数量和重复率随亮度变化而急剧变化,针对该问题,本文提出了一种基于局部自适应阈值的特征点提取算法。该算法通过设置自适应参数,用动态局部阈值计算每一个像素阈值来筛选特征点,解决了单一阈值选取不当导致的特征点丢失或重叠问题,对像素亮度进行分类,逐层筛选出候选特征点,达到定向二进制简单描述符特征点的精确提取。实验结果表明:亮度在增加或减少60%范围内变化时,特征点分布均匀无重叠,特征点数量的极差为80,仅占原始亮度特征点数量的25%,整体重复率稳定在80%以上。

In environmental perception of outdoor scenes,the number and repetition rate of feature points extracted based on fixed threshold change dramatically with the change of brightness.To solve this problem,a feature point extraction algorithm based on local adaptive threshold was proposed.The algorithm filtered feature points by setting adaptive parameters and calculated each pixel threshold with dynamic local threshold. The feature points loss or redundancy caused by improperly selection of fixed threshold were overcomed. By classifying the brightness of pixels,the candidate feature points were screened layer by layer to achieve accurate extraction of oriented fast and rotated brief( ORB) feature points.The experimental results show that when the brightness decreases or increases within 60%,the distribution of feature points is uniform without overlap,and the range of the number of feature points is only 80,accounting for only 25% of feature points of original brightness.The overall repetition rate is stable at more than 80%.