基于加权霍夫投票的多视角车辆检测方法

Multi-view vehicle detection method based on weighted Hough voting

针对复杂场景中车辆由于视角变化引起的检测精确度过低的问题,改进霍夫投票目标检测模型,提出一种在统一框架下通过不同权重组合发现目标最优视角并进行精确定位的方法。首先,利用一种无监督方法实现多视角车辆的子视角划分;其次,利用子视角划分结果定义霍夫投票过程中各正例样本在不同视角下的投票权重;最后,利用子视角划分和投票权重,提出一种新的适用于多视角目标检测的加权霍夫投票模型。在MITStreet Scene Cars和PASCAL VOC2007 Cars两个常用数据集上的实验结果表明,所提方法在不增加模型复杂度的前提下,有效提升了多视角目标检测精确度。

For the low detection accuracy of vehicles in complex scene caused by view variation,the target detection model based on Hough voting is improved to propose an accurate location method to find the object optimal view by means of different weights combination in the unified framework.An unsupervised method is used to realize the sub-view division for the multi-view vehicles.According to the sub-view division result,the voting weight of each positive example under different views is defined in the process of Hough voting.The sub-view division and voting weight are used to propose a weighted Hough voting model suitable for multi-view target detection.The experimental results are obtained by two commonly-used datasets of MITStreetScene Cars and PASCAL VOC2007 Cars.The experimental results demonstrate that the method can improve the multi-view target detection accuracy without increasing the model complexity.