Research Of Constant Color Correction Algorithms For Froth Image Based On Image Statistical Modeling

In the process of mineral flotation monitoring based on machine vision, the color information of bubble image can directly reflect the stand or fall of flotation conditions. Therefore, the optimal control of flotation conditions can be realized by extracting and studying the color feature of the flotation foam image. However, due to the bad environment and the complex illumination of the flotation scene, the froth image always suffers serious noise and color cast. In order to eliminate the above adverse effects and improve the veracity of color feature extraction, color correction algorithms for froth image is proposed based on image statistical modeling theory in this paper. The main research contents of this paper are organized as follows:(1) In view of the serious noise in flotation bubble picture, a morphology denoising method is presented based on constant color space model according to the visual characteristics of froth image in this paper. This method transforms the froth image to a special RGB model in which color cast is removed, in order to eliminate the excursion of image histogram led by illumination variation. Then, the recombination morphology filter is used to remove the noise of froth image and the filter is designed with multiple structural elements on the basis of froth image’s shape feature. With the purpose of reducing the loss of image details during filtering, Top-hat transformation is applied to froth image at the same time. At last, the effectiveness of the proposed algorithm is verified by simulation analysis, segmentation effect after denoising analysis and calculation of the peak signal to noise ratio. This algorithm not only removes the noise well but also retains the color information and details such as edges or textures of bubble image. A good foundation for the follow-up feature extraction can be laid by the algorithm as well.(2) In order to reduce the flotation field complex illumination influence on color information of bubble image, a froth image color correction algorithm is propsed in this paper based on image statistical model in Contourlet domain. At first, through the analysis of the statistical characteristics of froth image, Contourlet transform and generalized Gaussian distribution are used to build statistical model of image and the froth image is eventually converted to vector form. Next, pictures with known real light in standard picture library of color constancy (Gray-ball) are chose as the training samples and their statistical parameters are calculated to compose a statistical model parameter set. Five common color constancy algorithms are selected to do color correction and the minimum angle error is used to mark the best color constancy algorithm of each picture. In the end, the best color constancy algorithm of original froth image is achieved by KNN. The experimental results shows that the algorithm can obtain a good color correction effect for froth image. Figures (45), tables (5), references (62).