원문정보
초록
영어
To meet the demand for the early location of fire in large-span space buildings, an accurate fire location method is proposed based on machine vision technology. A nonlinear implicit camera calibration method is proposed by combining an improved particle swarm optimization (PSO) method with least squares support vector machine (LS-SVM) to solve the problem that it is difficult to establish accurate mathematical models for traditional nonlinear explicit camera calibration. The matched pixel coordinates of images collected by cameras are used as input, and the output is the world coordinates. The IPSO is used to search the optimal parameters of LS-SVM regression model to increase the convergence speed and improve the generalization ability of LS-SVM. The spatial location of fire is achieved by three-dimensional reconstruction. The proposed method is applied to fire location for high and large buildings, and experimental results show that the method is effective, fast and accurate.
목차
1. Introduction
2. Basic Theories
2.1 Camera Imaging Model
2.2 Improving PSO
2.3 LS-SVM Theory
2.4 IPSO for LS-SVM Parameter Optimization
2.5 LS-SVM Prediction Model based on IPSO
3. Experimental Analysis
3.1 Calibration Experiment
3.2 Calculation of 3D Position of Fire Source
5. Conclusions
References