Research Of Smoke Removal System For Image In Fire Scenario

According to some statistics, there are about 6~7 million fire accident on a global scale every year, it causes about 300,000 people’s death. Nowadays, drones have been used to observe fire situations, and rescue workers can find out the better ways. But the fire smoke greatly affects the rescue operation. It makes relief more difficulty, it may cause more damage. The molecular in fire smoke produce scattering effect, it affects the visual effects of rescue workers. So do the research of smoke removal is very improtant and meaningful. The rescue workers could decide how to rescue more effectly by the image of smoke removal.This paper constructs the self-adaption smoke removal system. First of all, do some image preprocessing and classify the images of fire scenario by illumination intensity. The algorithm based on retinex theory is more suitable for bright photots;the image de-haze algorithm based on dark channel prior is more suitable for dark photots. And optimize the two algorithms.Because the colour of images is dark, so this papaer improves the brightness and contrast of images by fuzzy image enhancement algorithme. The images are divided into four classes:outside during daytime, indoors during daytime, indoors during nights, outside during nights. The images of outside during daytime is more bright, so the algorithm based on retinex theory is better.The other three kinds of images is darker, so the image de-haze algorithm based on dark channel prior is preferred. This paper puts forward the algorithm of image classification.This paper has done the research into the image de-haze algorithm based on dark channel prior. Because the low efficiency of transmissivity algorithm, the computing efficiency is very low.So reduce the sample of atmospheric transmittance and then use the interpolation method to slove transmissivity of original images for the problem of excessive calculation and low computing efficiency; By splitting the bright area such as sky area to correct the transmissivity.For solving the problem of wrong transmittance of bright area in de-haze algorithm based on dark channel prior. This paper puts forward the revalidation of uncertain sky area and makes sure weather the area is belong to the sky area, it guaranteess the detection right, and the result of experiment testifies it can prevent over exposing of sky area; The image after disposed by de-haze algorithm based on dark channel prior will beome dim. This paper puts forward the color enhancement algorithm to do the adjustment of contrast. And the results of experiment prove the solution can enhance the brightness of color and the contrast of image.The improved algorithms has been verified to optimize the efficiency and ensure the effect of images’ smoke removal. To slove the inefficiency of retinex theory, this paper proposes to use laplace to detect the edges,and do brightness enhancing for the V channel in HSV space, it improves the algorithm effectively,and the effect of smoke removal is very good.This paper classifies the image of fire scenario, it constructs the adaptive system of images’ smoke removal. It improves the algorithms, and implements the adaptive system of smoke removal. According to experiments,this paper verifies the systems’ efficiency and applicability.