改进的AdaBoost-SVM算法用于无人机目标跟踪

Improved AdaBoost-SVM Algorithm for UAV Target Tracking

研究一种基于AdaBoost的支持向量机(SVM)用于无人机的目标检测与跟踪.给出了无人机目标跟踪的基本框架和HOG特征提取的步骤,分别总结了AdaBoost算法与SVM的分类原理以及将基于径向基核函数的支持向量机(RBFSVM)作为AdaBoost的弱分类器时,σ值的取值大小对AdaBoost分类器的影响问题,给出目标跟踪算法的计算流程.利用人脸识别标准库ORL验证了随机投影方法进行数据降维的有效性.由两种数据集的分类实验结果以及最终跟踪结果的精确度曲线与成功率曲线表明,与传统的跟踪算法相比,提出的基于机器学习的目标跟踪算法可有效处理无人机目标跟踪中的光照变化、目标遮挡、运动模糊和小目标检测等问题.

A support vector machine(SVM)based on AdaBoost was studied for UAV target detection and tracking.The basic framework of UAV target tracking and the steps of HOG feature extraction were given.The classification principles of AdaBoost algorithm and SVM were summarized respectively.The influence ofσvalue on AdaBoost classifier was also discussed when RBFSVM based on radial basis function was used as AdaBoost weak classifier.Finally,the calculation flow of the target tracking algorithm was given.The ORL of face recognition standard library was used to verify the effectiveness of the random projection method for data dimensionality reduction.The classification experiment results of the two data sets and the accuracy curve and success rate curve of the final tracking results show that compared with the traditional tracking algorithm,the proposed target tracking algorithm based on machine learning can effectively deal with the problems of illumination change,target occlusion,motion blur and small target detection in UAV target tracking.