CAJ | 학술논문

基于视觉的自主火灾检测在火灾预警方面有很重要的应用价值。该文提出了一种基于水平集的火灾检测方法，通过准确分割图像中的火焰部分来识别火灾。该方法将火灾方面的特定信息集成到水平集框架中：首先得到火灾的概率观测模型，然后通过考虑最大后验概率和边缘线索形成水平集的能量函数，最后通过最小化能量函数，使用水平集的轮廓能准确地划清火焰的区域。算法在一个含有各种火灾场景的综合性图像集上进行了测试，实验结果验证了算法的有效性。
기우시각적자주화재검측재화재예경방면유흔중요적응용개치。해문제출료일충기우수평집적화재검측방법，통과준학분할도상중적화염부분래식별화재。해방법장화재방면적특정신식집성도수평집광가중：수선득도화재적개솔관측모형，연후통과고필최대후험개솔화변연선색형성수평집적능량함수，최후통과최소화능량함수，사용수평집적륜곽능준학지화청화염적구역。산법재일개함유각충화재장경적종합성도상집상진행료측시，실험결과험증료산법적유효성。
Although level set methods are widely used in object detection and segmentation , they are seldom applied to fine detection in literatures .In this paper , it presents a level set based method for fire detection , especially , for precise flame segmentation which is essential in fire metrology .The proposed method incorpo-rates fire specific knowledge into the level set framework .In particular , it first learns probabilistic observation models of fire and background from training data .Then the level set energy functional is formulated by consider-ing both the maximization of the energy functional , the obtained level set contours can precisely delimit flame regions.The proposed algorithm is tested on a comprehensive image set containing diverse fire scenarios .The experimental results validate the effectiveness of our algorithm .