A novel dominant correlogram based particle filter was proposed for an object tracking in visual surveillance. Particle filter outperforms the Kalman filter in non-linear and non-Gaussian estimation problem. This pape...A novel dominant correlogram based particle filter was proposed for an object tracking in visual surveillance. Particle filter outperforms the Kalman filter in non-linear and non-Gaussian estimation problem. This paper proposed incorporating spatial information into visual feature, and yields a reliable likelihood description of the observation and prediction. A similarity-ratio is defined to evaluate the effectivity of different similarity measurements in weighing samples. The experimental results demonstrate the effective and robust performance compared with the histogram based tracking in traffic scenes.展开更多
文摘A novel dominant correlogram based particle filter was proposed for an object tracking in visual surveillance. Particle filter outperforms the Kalman filter in non-linear and non-Gaussian estimation problem. This paper proposed incorporating spatial information into visual feature, and yields a reliable likelihood description of the observation and prediction. A similarity-ratio is defined to evaluate the effectivity of different similarity measurements in weighing samples. The experimental results demonstrate the effective and robust performance compared with the histogram based tracking in traffic scenes.
文摘以交叉相关光谱匹配(cross correlogram spectral matching,CCSM)为基础构建土地覆盖变化强度指标,利用华北农业植被覆盖区2期不同时相的TM图像计算该地区土地覆盖变化强度图像。认为变化强度图像任意二阶邻域中像素的变化强度服从隐马尔可夫模型,用马尔可夫随机场-最大后验估计(maxium a posteriori estimation of markovrandom field,MRF-MAP)的方法从变化强度图像中提取植被变化区域。实验证明:该方法能够有效识别各种外源噪声造成的农业植被覆盖区域同物异谱的现象,可准确提取植被变化区域;但对于水体区域存在误判现象。
