CAJ | 학술논문

在多红外火焰探测系统中，提出了一种基于决策树的火灾识别算法。按照特种火灾探测器的国家标准实验的要求，获取实验数据。该算法首先对五个红外火焰探测器获得的数据进行多窗口重叠交叉预处理，然后提取六个火灾特征作为决策树的分类属性，对决策树进行训练、剪枝，最后得到火灾识别的最优决策树模型。将该识别模型应用于在线火灾识别，实验结果表明该决策树分类算法的准确率可以达到95.2%，识别速度在2s以内，较其他的分类识别算法有更高的准确率和更快的识别速度，具有很好的实用性。
재다홍외화염탐측계통중，제출료일충기우결책수적화재식별산법。안조특충화재탐측기적국가표준실험적요구，획취실험수거。해산법수선대오개홍외화염탐측기획득적수거진행다창구중첩교차예처리，연후제취륙개화재특정작위결책수적분류속성，대결책수진행훈련、전지，최후득도화재식별적최우결책수모형。장해식별모형응용우재선화재식별，실험결과표명해결책수분류산법적준학솔가이체도95.2%，식별속도재2s이내，교기타적분류식별산법유경고적준학솔화경쾌적식별속도，구유흔호적실용성。
In the multi-infrared flame detection,a fire recognition algorithm based on decision tree is proposed. According to the National Standard for Special Fire Detectors,large number of experimental data are acquired. Firstly,the acquired data of the five infrared flame de-tector are pretreated by the overlapping cross way in the algorithm. Then six characteristics of fire are extracted as a decision tree classified attributes,and decision tree is trained and pruned. Finally,the optimal decision tree model for fire detection is obtained. This recognition model is applied to the online fire detection,the experimental results show that the accuracy of the decision tree classified algorithm can a-chieve 95. 2% and the recognition speed is less than 2s. Compared with other recognizable algorithms,decision tree has higher accuracy and faster recognition speed. It is of great practicality.