摘要
为了能够提升视频技术下车辆检测的正确率,论文提出结合使用HOG特征与SIFT特征作为车辆检测的特征提取算法,再通过支持向量机(SVM)将样本数据划分为训练集与验证集,使用不同核函数进行训练和验证,确定最优核函数为高斯核函数。最后将训练的模型使用到视频文件进行车辆的预测。最终,实验数据表明,该方法提升了传统的HOG+SVM的样本检测效率,高斯核函数下检测率高达98.38%。处理视频文件时车辆检测效果良好,但是模型仍不够稳定,希望日后完善训练集继续改进算法。
In order to improve the accuracy of vehicle detection under video technology,this paper proposes a combination of HOG and SIFT features as an extraction algorithm for vehicle detection features,and then the sample data is divided into a training set and verification set through support vector machine(SVM).Different kernel functions are trained and verified to determine the optimal kernel function as a Gaussian kernel function.Finally,the trained model is applied to the video files for vehicle prediction.Finally,the experimental data show that the proposed method improves the sample detection efficiency of the traditional HOG+SVM,and the detection rate under the Gaussian kernel function is as higher as 98.38%.Vehicle detection results are good when pro⁃cessing video files,but the model is still not stable enough,and it is hoped that the training set will continue to improve the algo⁃rithm in the future.
作者
关晓斌
李战明
GUAN Xiaobin;LI Zhanming(College of Electrical Engineering and Information Engineering,Lanzhou University of Technology,Lanzhou 730050)
出处
《计算机与数字工程》
2021年第6期1113-1117,共5页
Computer & Digital Engineering