基于多源传感器矿井移动目标跟踪与定位.docVIP

基于多源传感器矿井移动目标跟踪与定位 　　摘要： 针对井下NLOS环境信号接收强度（RSSI）和飞行时间（ToF）定位方法存在多径干扰和定位时延，导致定位误差较大问题，提出了基于多源传感器的矿井移动目标跟踪与定位方法。首先通过建立矿井移动目标跟踪与定位模型，利用混合卡尔曼粒子滤波算法监测井下移动目标的位置范围，在阈值内触发CCD视觉传感器对移动目标图像信息采样和特征提取。将预测和估计的移动目标坐标转化为CCD视觉传感器的图像物理坐标系坐标，并进行特征匹配图像的分割，以缩小CCD视觉传感器的特征匹配范围，提高匹配速度。最后利用SIFT算法对已训练图像与采集的目标图像进行特征匹配并对移动目标进行融合定位，根据CCD视觉传感器所在基站位置获取和校正矿井移动目标的准确位置信息，实现实时跟踪与精确定位。实验表明，与其他算法相比，在视觉传感器监测阈值内，本文方法能够有效提高定位精度和实时性，对井下目标遮挡、低照度和噪声环境下移动目标跟踪与定位具有较强的鲁棒性。 　　关键词： 矿井；混合卡尔曼粒子滤波算法；SIFT算法；CCD视觉传感器；跟踪与定位 　　中图分类号：TP311 文献标识码：A 文章编号：1009-3044（2018）09-0252-04 　　Abstract： For underground NLOS environment， signal reception intensity （RSSI） and time of flight （ToF） location methods exist multipath interference and location delay， resulting in large positioning error. A mine moving target tracking and location method based on multi-source sensor is proposed. First， a mine moving target tracking and positioning model is established. The hybrid Calman particle filter algorithm is used to monitor the location range of the underground moving target， and trigger the CCD vision sensor to sample and extract the moving target image information within the threshold range. The predicted and estimated coordinates of the moving target are transformed into the coordinates of the image physical coordinate system of the CCD vision sensor， and the feature matching image is segmented so as to reduce the feature matching range of the CCD vision sensor and improve the matching speed. Finally， the SIFT algorithm is used to match the features of the trained image and the acquired target image， and the hybrid Kalman particle filter algorithm is used to fuse the moving target. So that the accurate position information of mine moving target can be acquired and corrected by using the location of the base station where the CCD vision sensor is located， so as to realize real-time tracking and precise positioning. Experiments show that compared with other algorithms， the proposed method can effectively improve location accuracy a