基于机器视觉的大空间建筑火源定位方法研究

发布时间：2018-12-12 19:02

【摘要】：随着国民经济迅速发展,大空间建筑日益增多，因其特殊性和复杂性，导致传统的火源探测和灭火设施的功能很难施展。近年来，图像型火灾探测技术已很好地应用在大空间建筑消防中，智能消防炮灭火技术也得到一定的应用，论文在研究传统智能消防炮火源定位方法的基础上，，针对其明显滞后性的缺点，采用基于平行双目立体视觉的火源定位方法，提高了火源定位的实时性和准确性，具有重要的应用价值。论文采用平行双目立体视觉和三维信息重建技术来确定大空间建筑火源的三维坐标。首先，用张氏平面标定法对摄像机内部参数进行定标；其次，用SURF算法提取火焰图像特征点，进行快速灰度相关粗匹配，然后估计基础矩阵以恢复对极几何关系，再用SURF算法进行特征点立体匹配。其中，为了提高定位的实时性和准确度，在基础矩阵估计时，采用了一种运算速度较快、准确度较高的先LMedS后M-Estimators的估计方法，即先用LMedS算法去除异常匹配点，再用M-Estimators算法去除定位噪声；在保持SURF匹配算法优点的基础上，对其匹配搜索策略改进，采用特征集分组，按组依次匹配的策略，提高了匹配的精度和时效性；最后，由摄像机内部参数和基础矩阵计算出本质矩阵，对本质矩阵进行奇异值分解得到平行双目摄像机相对位置的外部参数，再利用三维重建技术，来确定火源三维深度信息。经仿真实验证明，论文的定位方法不仅能保证火源定位的实时性要求，还可以提高火源定位的精确度，可以满足大空间建筑火源空间定位的要求。
[Abstract]:With the rapid development of national economy, the large space building is increasing day by day, because of its particularity and complexity, the function of traditional fire source detection and fire extinguishing facilities is very difficult to carry out. In recent years, image fire detection technology has been well applied in large space building fire fighting, and intelligent fire extinguishing technology has also been applied to a certain extent. Aiming at its obvious lag, the fire source location method based on parallel binocular stereo vision is adopted, which improves the real-time and accuracy of fire source location, and has important application value. In this paper, parallel binocular stereo vision and three-dimensional information reconstruction technology are used to determine the three-dimensional coordinates of the fire source of large-space buildings. Firstly, the camera internal parameters are calibrated by Zhang's plane calibration method. Secondly, the feature points of flame images are extracted by SURF algorithm, and the coarse matching of fast gray correlation is carried out, then the basic matrix is estimated to restore the relations between polar geometry, and the stereo matching of feature points is carried out using SURF algorithm. In order to improve the real time and accuracy of localization, a fast and accurate estimation method of LMedS and M-Estimators is adopted in the estimation of basic matrix, that is, LMedS algorithm is used to remove the abnormal matching points. Then the M-Estimators algorithm is used to remove the location noise. On the basis of keeping the advantages of SURF matching algorithm, the matching search strategy is improved, and the matching accuracy and timeliness are improved by grouping feature sets and matching according to group sequence. Finally, the essential matrix is calculated from the camera internal parameters and the basic matrix, and the external parameters of the relative position of the parallel binocular camera are obtained by singular value decomposition of the essential matrix. The 3D depth information of the fire source is determined by using the three-dimensional reconstruction technique. The simulation results show that the method can not only meet the real-time requirements of fire source location, but also improve the accuracy of fire source location, and can meet the requirements of fire source location in large space buildings.
【学位授予单位】：西安建筑科技大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TU892;TP391.41

【参考文献】