CAJ | 학술논문

针对快速SIFT（Scale Invariant Feature Transform）算法去掉高斯加权环节，导致算法稳定性下降的不足，对快速SIFT算法中的特征描述子进行了改进，用圆形代替矩形并对区域进行分层，越靠近关键点层数越多，相当于对中心区域进行了加权，并通过仿真给出了最佳加权层数和特征描述子方向数。仿真实验表明：该算法不但满足了匹配正确率和实时性的要求，而且具有很高的稳定性和鲁棒性，能够满足目标跟踪的实时性要求。
침대쾌속SIFT（Scale Invariant Feature Transform）산법거도고사가권배절，도치산법은정성하강적불족，대쾌속SIFT산법중적특정묘술자진행료개진，용원형대체구형병대구역진행분층，월고근관건점층수월다，상당우대중심구역진행료가권，병통과방진급출료최가가권층수화특정묘술자방향수。방진실험표명：해산법불단만족료필배정학솔화실시성적요구，이차구유흔고적은정성화로봉성，능구만족목표근종적실시성요구。
In the fast SIFT (scale invariant feature transform)algorithm,it removes Gaussian weighted links algorithm and the stability is decreased.For that question,we improved the feature descriptor of fast SIFT algorithm.In the improved algorithm,we stratified the area with circular instead of rectangular. Therefore,the closer to the key points,the more layers are.The algorithm which was equivalent to the central region was weighted,and the optimum number of layers and feature descriptor weighted number of directions through simulation were given.The simulation results show that the method not only content the request for correct matching rate and real time,but also have high stability and robustness which can meet the real time demand of target tracking.