Title :
Adaptive flame image detection algorithm
Author :
Yang, Najuan ; Wang, Huiqin ; Zhang, Qianyuan ; Ma, Zongfang
Author_Institution :
Sch. of Inf. & Control Eng., Xi´´an Univ. of Archit. & Technol., Xi´´an, China
Abstract :
To overcome the disadvantage of the traditional fire detect methods in large space buildings, a new flame image detection algorithm based on Support Vector Machine (SVD) was studied in this paper. SVD is an effective data classification tool and it can efficiently classify the fire flames from the suspicious source. In order to improve the fire recognition accuracy, an adaptive image fire detection algorithm is researched. The core idea of the algorithm is: considering the influence of sample data when selecting the kernel function, constructing the modified kernel function with conformal transformation method of information geometry, training SVM by the modified kernel function. The experiment results show that this algorithm improved the classification precision and the recognition accuracy of fire image.
Keywords :
adaptive signal processing; building; flames; image recognition; support vector machines; adaptive flame image detection; conformal transformation method; kernel function; large space buildings; support vector machine; Classification algorithms; Equations; Feature extraction; Fires; Kernel; Mathematical model; Support vector machines; adaptie algorithms; feature extraction; fires; image classification; support vector machine;
Conference_Titel :
Image and Signal Processing (CISP), 2010 3rd International Congress on
Conference_Location :
Yantai
Print_ISBN :
978-1-4244-6513-2
DOI :
10.1109/CISP.2010.5647977