大空间建筑火源的精确定位方法

摘要

针对大空间建筑火源定位精度低和实时性差的问题，提出一种基于火焰图像熵值加速鲁棒性特征 SURF（Speeded Up Robust Features）的改进遗传算法求解框架应用在大空间建筑火源定位中。在良好的火源探测下，首先用 SURF 算法提取火焰特征点，引入特征点熵值来减少误匹配，得到较好的匹配点对，进一步求解摄像机参数。通过改进选择、交叉、变异算子设计大空间建筑火源定位中摄像机标定的遗传算法，最终获得火源三维信息。实验结果表明，提出的新算法应用在大空间建筑火源定位中能更好恢复火源的三维深度信息。%In order to solve the problems of low accuracy in fire source localisation and poor real-time property in large-space buildings,we proposed a solution framework of improved genetic algorithm,which is based on flame image entropy accelerating the robust feature (SURF), to apply in the fire source localisation of large-space buildings.Under the condition of good fire source detection,it first extracts flame feature points with SURF algorithm,then introduces the entropy of the feature points to reduce mismatching,and gets better matching-point pairs to further solve the parameters of video camera.We designed a genetic algorithm for camera calibration in fire source localisation of large-space buildings by improving the selection,crossover and mutation operators,and finally obtained the three-dimensional information of fire source. Experimental result illustrated that applying the proposed new algorithm in fire source localisation of large-space buildings can better restore 3D depth information of fire source.