Homogenous fog condition recognition based on traffic scene

Light transition model in homogenous fog has extensive applications. Nowadays it is used to estimate scene distance and used for 3-D reconstruction especially in traffic scene. This is because the whole road could be identified as uniform. However, the foundation requite for these application is the current weather condition is the homogenous fog weather condition which is presupposed and not identified by most of the authors. A methodology for homogenous fog weather condition recognition for transport scene based on texture analysis and light transition model is presented in this paper. Firstly, a narrow area containing the sky and road is extracted based on texture analysis. This narrow area should be consistent from the top to bottom of scene background. Secondly, mean value for each line is calculated and saved. We call these data Mean Pixel Values (MPV). These values are sorted from the top to the bottom and fitted by mathematics function. Finally, the mathematic function style is compared with exponential function. If the mathematic function we calculated is an exponential function, then the current weather condition is homogenous fog condition, otherwise, other weather conditions. We test our algorithm by using traffic scene in four weather conditions. Experimental results verify the effectiveness of the algorithm.