【摘要】 在存在壁面反射的低照度火灾环境中,传统的火焰分割算法如颜色分割、运动检测等,在进行火焰分割时造成过分割现象,分割的效果不理想,影响后续的火灾正确识别。针对上述问题,提出了一种基于自动种子区域生长(Automatic Seeded Region Growing,ASRG)的火焰分割算法。首先将从火灾视频中获取的火灾图像从RGB颜色空间转换到YCbCr颜色空间,在Y通道中采用较大自适应阈值背景减法将火灾图像二值化,分别将可疑火焰像素点的横坐标和纵坐标按大小进行排序,取排序后的中间值作为种子点,再由原RGB火灾图像转换而成的灰度图像中,以该种子点进行区域生长,最后将区域生长后的火焰分割图像与采用较小自适应阈值背景减法得到的火焰分割图像进行交集处理,得到最终的火焰分割图像。实验表明ASRG算法在存在壁面反射的低照度火灾环境中,火焰分割效果好,有效解决了该环境下的火焰过分割问题,同时在其他火灾环境中也有较好的火焰分割效果。更多还原

【Abstract】 In the low-light fire environment with wall reflection,traditional flame segmentation algorithms such as color segmentation and motion detection have over-segmentation problem during flame segmentation.The fire segmentation does not work well and influences the subsequent fire recognition.To solve this problem,we propose a flame segmentation algorithm based on automatic seeded region growing(ASRG).First,the fire video images are converted from the RGB color space to the YCbCr color space,and binary images are obtained by background subtraction with a high adaptive threshold in the Y channel.Then,the horizontal and vertical coordinates of the suspected flame pixels are sorted according to their size,and the sorted intermediate value is used as the seed point.Finally,the final flame is obtained by the intersection of the binary image obtained by region growing with the seed point in the original gray fire image and that obtained by background subtraction with a low adaptive threshold in the Y channel.Experiments show that the proposed algorithm works well in the flame segmentation of low-light fire environment with wall reflection.Besides,the algorithm also performs well in other fire environments.更多还原