基于Stixel world及特征融合的双目立体视觉行人检测

针对单目视觉行人检测无法获得深度信息从而导致冗余信息较多、检测效率和准确度存在局限性的问题,首先,在图像的预处理阶段提出了一种利用双目立体视觉产生的视差信息优化分析来简化复杂场景的动态规划棒状像素场景(stixel-world)表达方式;然后,在行人目标检测阶段,对传统HOG特征中block尺度进行分析、降维,采用Fisher准则筛选得到了适用于道路环境下的多尺度HOG(multi-HOG)特征,将Multi-HOG特征与LUV颜色通道特征进行融合,最后采用交叉核支持向量机(hikSVM)分类器对行人目标分类。实验结果表明,采用改进过后的Stixel-world算法用于图像预处理极大地减少了计算时间。缩小了行人检测的候选区域,基于特征融合和hik-SVM的目标检测算法在保证检测准确度的前提下,具有较好的实时性和鲁棒性。

Stereo visual pedestrian detection based on Stixel world and feature fusion

The monocular vision pedestrian detection can not obtain the depth information, so detection efficiency and accuracy are limited. Firstly, in the image pre processing, the disparity information optimization analysis is proposed to simplify the expression of dynamic programming stixel world in complex scenes based on stereo vision. Then, at the stage of pedestrian target detection, this paper analyzes the influence of the block scale on the detection effect in the traditional HOG feature, and obtains the multi HOG feature which is suitable for the road environment using the fisher criterion. The multi HOG feature is integrated with the LUV color channel feature. Finally, the hik SVM is used for pedestrian target classification. The experimental results show that the improved Stixel world algorithm for image preprocessing greatly reduces the computation time and reduces the candidate region of pedestrian detection, target detection algorithm based on feature fusion and hik SVM has good real time and robustness under the premise of guaranteeing the detection accuracy.