CAJ | 학술논문

　　针对现有动态火焰识别方法中存在的经验阈值难以确定和因彩色信息丢失导致识别不准确等问题，本文提出一种奇异值分解特征提取配合支持向量机细化分类的火焰图像识别方法。该方法选用火焰的纹理信息作为奇异值分解的一个基础，运用支持向量机的方法对纹理中的干扰因素细化，运用标准FCM识别算法对火焰进行识别。实验结果表明，该算法可排除高亮区域的干扰，准确识别出火焰区域，火焰的识别准确度较高。
　　침대현유동태화염식별방법중존재적경험역치난이학정화인채색신식주실도치식별불준학등문제，본문제출일충기이치분해특정제취배합지지향량궤세화분류적화염도상식별방법。해방법선용화염적문리신식작위기이치분해적일개기출，운용지지향량궤적방법대문리중적간우인소세화，운용표준FCM식별산법대화염진행식별。실험결과표명，해산법가배제고량구역적간우，준학식별출화염구역，화염적식별준학도교고。
Existing dynamic flame identification method of the experience of existing threshold and is difficult to deter-mine because of color information loss lead to inaccurate recognition and so on, this paper presents a kind of singular value decomposition feature extraction with support vector machine refining classification flame image recognition method. This method is used for flame texture information as a singular value decomposition of a foundation, using sup-port vector machine method to texture the interference factors refining, using standard FCM algorithm to identify flame recognition. The experimental results show that this algorithm can eliminate the interference of highlight areas, accurately identify flame area in flames, and the recognition accuracy is higher.