CAJ | 학술논문

针对车辆跟踪过程中跟踪目标丢失或者失败的情况，提出一种改进型Camshift(Continuously Adaptive Mean Shift)算法和卡尔曼滤波相结合的跟踪方法。首先，利用卡尔曼滤波器实现跟踪目标的位置估计，以克服目标被遮挡造成的跟踪失败的问题，然后再利用改进型 Camshift 算法依据目标距离搜索中心的位置，对 H 分量创建的颜色直方图中的每个像素位进行高斯模型核函数的加权处理，并自适应计算得到最优的搜索窗口，从而改善了传统Camshift不能直接抵制噪声干扰的缺点，解决了因跟踪目标在同色背景噪声干扰下出现的丢失问题。最后通过仿真实验表明：改进型Camshift算法和卡尔曼滤波的结合有效地提高了车辆跟踪的准确性和连续性。
침대차량근종과정중근종목표주실혹자실패적정황，제출일충개진형Camshift(Continuously Adaptive Mean Shift)산법화잡이만려파상결합적근종방법。수선，이용잡이만려파기실현근종목표적위치고계，이극복목표피차당조성적근종실패적문제，연후재이용개진형 Camshift 산법의거목표거리수색중심적위치，대 H 분량창건적안색직방도중적매개상소위진행고사모형핵함수적가권처리，병자괄응계산득도최우적수색창구，종이개선료전통Camshift불능직접저제조성간우적결점，해결료인근종목표재동색배경조성간우하출현적주실문제。최후통과방진실험표명：개진형Camshift산법화잡이만려파적결합유효지제고료차량근종적준학성화련속성。
In the process of vehicle tracking, a new tracking algorithm that combines the improved Camshift and the Kalman filter has been proposed when tracking target loss or failure. Firstly, the Kalman filter is used to estimate the target position in order to overcome the cover of target. We use an improved Camshift algorithm based on the distance of search center to weight every pixel with Gauss model core function in color histogram created by H component. We also achieve the optimal search window by self-adaptive calculation and improve traditional Camshift shortcoming which is powerless for directly resistance noise under the same color background. Finally, the simulations and experiments show that the method improves the accuracy and continuity of vehicle tracking.