火灾视觉特征的提取是视觉火灾探测中的关键问题. 我们主要研究色彩、纹理以及轮廓脉动

等特征的提取,并提出一种度量轮廓脉动信息的距离模型,该模型在规格化的傅立叶描述子空间能

够准确地度量这种时空闪烁特征. 实验结果表明,该方法具有比较好的鲁棒性,有助于提高视觉火

灾探测的准确率、降低误报漏报率.-Based on investigating color , text ure and temporal feat ures for vision based fire detection , a

distance model of contour fluct uation between two successive f rames in t he normalized Fourier descr iptor s

domain was presented to measure t his time varying contour fluct uation feat ure of flame. The model of

contour fluct uation is effective and robust for fire recognition. To f urt her reduce fal se alarms , several

features ext racted according to color , text ure and the distance model were toget her regarded as a joint

feature vector for artificial neural network to detect fire. Experiment s show t hat the algorithm is effective

and robust , and t hat it is significant for improving accuracy and reducing fal se alarms.