基于逐步判别法与BP神经网络的火焰目标检测  

FLAME TARGET DETECTION BASED ON STEPWISE DISCRIMINATION AND BP NEURAL NETWORK

作　　者：刘敏 潘新[1] 刘菲 周艳青[1] 姜坤 LIU Min;PAN Xin;LIU Fei;ZHOU Yanqing;JIANG Kun(College of Computer and Information Engineering,Inner Mongolia Agricultural University,Hohhot 010018,China;Hohhot Fire Monitoring Management Center,Hohhot 010010,China;Baotou Service Management Vocational School,Baotou 014060,China)

机构地区：[1]内蒙古农业大学计算机与信息工程学院,呼和浩特010018 [2]呼和浩特市火灾监控管理中心,呼和浩特010010 [3]包头服务管理职业学校,包头014060

出　　处：《内蒙古农业大学学报（自然科学版）》2021年第4期92-96,共5页Journal of Inner Mongolia Agricultural University(Natural Science Edition)

基　　金：国家自然科学基金项目(61562067,61962048);全国农业专业学位研究生教育指导委员会教学改革研究项目(2019-NYYB-26);中国高校计算机教育MOOC联盟线上线下混合式教学改革项目(B190140);内蒙古农业大学研究生重点课程建设项目《应用数理统计》。

摘　　要：在基于图像分析的火焰目标检测判别函数中,判别特征的选取会显著影响判别的准确率,为了提高检测的准确率并且降低火焰目标特征向量的维数,本文提出了基于逐步判别法与BP神经网络的火焰目标检测方法,该方法对火焰目标的颜色、亮度和纹理特征及其检测算法进行了研究,然后利用逐步判别法筛选出区分能力强的特征向量子集并作为BP神经网络的输入端分量完成火焰目标的识别,该方法的漏报率为6.7%,准确率为93.3%,火焰目标检测效果最佳。In the discriminant function of the flame target detection based on the image analysis,the selection of discriminant features will significantly affect the discriminant accuracy.In order to improve the detection accuracy and reduce the dimension of the flame target feature vector,a flame target detection method based on the stepwise discriminant method and BP neural network was proposed in this paper.Through this method,the color,brightness,texture characteristics of the flame target and its detection algorithm were studied,and then the feature vector subsets with strong discrimination abilities were selected by the stepwise discriminant method,which were used as the input components of the BP neural network to complete the recognition of the flame target.The false alarm rate of this method was 6.7%,the accuracy rate was 93.3%,and the detection effect of the flame target was the best.

关 键 词：火焰目标检测 特征向量 逐步判别 BP神经网络 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术] X932[自动化与计算机技术—计算机科学与技术]