Image Dehazing Algorithm For Iron Ore Green Pellets

Iron ore pellets are very important agglomerated rich ore,which plays an important role in steel industry.Manufacturing of green pellets in disk pelletizer is one of the important process in the production line of iron ore pellets.The quality of green pellets(such as pellet size)directly affects the physical properties of end products.With the development of industrial automation and the application of image technology,the online monitoring of the green pellets quality is mostly achieved by using machine vision.However,the image quality can be seriously deteriorated due to the harsh production environment such as haze and dust,which might result in failure in image processing algorithm and measuring errors.Current image dehazing algorithms focus mainly on images under natural illumination and environment,and seldom consider industrial images under artificial illumination conditions.Hence,by analyzing the features of green pellets images,this paper proposed a suitable dehazing method to remove haze in green pellets images,which is of theoretical importance and meaningful for practice.In this paper,the imaging of green pellets is studied,and then the traditional atmospheric scattering model is improved,based on which a novel dehazing algorithm based on fusing dark channel image is proposed;Finally,the proposed algorithm is verified by haze images collected from laboratory and steel company.The performance of the proposed algorithm is compared with current image dehazing algorithms.The main work and conclusions of this paper are as follows:(1)A platform generating haze images was established in laboratory to simulate the production environment of green pellets.With this platform,pellet images under different illuminations,different haze concentrations and morphologies can be captured.In the present work,more than 3000 images were collected,which laid a good basis for the study of image dehazing algorithms.(2)The imaging process of green pellets in hazy environment is studied,and then the traditional atmospheric scattering model is improved by modifying the direct attenuation term to the secondary attenuation term.The improved atmospheric scattering model is more suitable for artificial illumination environment.With it,a new image dehazing algorithm based on image fusion is designed.The dark channel images of multiple scales are firstly obtained,then the high frequency components are removed by wavelet decomposition while the low frequency components are fused by local variance fusion rules to obtain the accuratetransmittance.Finally,the transmittance is optimized by gradient guided filter and then substituted into the improved scattering model to restore the haze-free image.(3)The hazy green pellets images collected from laboratory were used to verify the proposed image dehazing algorithm.It shows good dehazing performance for pellet images under different illuminations,haze concentrations and morphologies,the restored images are very close to the corresponding reference images.Compared with the current image dehazing algorithms,the proposed algorithm has obvious advantages in terms of indices including peak signal-to-noise ratio and structural similarity.(4)Green pellets images with different haze concentration that were captured continuously in a local steel company were used to verify the proposed image dehazing algorithm.Results show that it can effectively solve the haze removal problem in industrial green pellet images.The sharpness of the restored images was significantly enhanced,and the processing speed of about 1.3 seconds per image,which can fully meet the real-time requirements of engineering application.