Research on low illumination video enhancement technology in coal mine heading face

Aiming at overcoming low illumination, uneven brightness, blurry texture and more noise in the video of coal mine heading face, a low illumination video enhancement algorithm was proposed for coal mine heading face. Firstly, the separability of convolution was utilized to carry out one-dimensional horizontal and vertical convolution of video images, then the perfect reflection method was used to achieve the automatic white balance, and the image hybrid enhancement technology was utilized to improve the overall brightness of the video images. Then, the image was divided into the highlight area, the middle tone area and the dark tone area by recursive segmentation based on the atmospheric scattering model and the dark channel prior method, and the maximum channel pixel of the corresponding interval was obtained. Besides, the mean value of the three maximum pixel values was taken as the estimation value of atmospheric illumination, and the transmittance was adjusted and optimized by introducing the adjustment factor. Meanwhile, the Laplacian sharpening process was used to increase the high frequency component and suppress the low frequency component of the image to increase the image contrast. Finally, the low-illumination video of heading face was dehazed based on the improved atmospheric scattering model. The experimental results show that the proposed video enhancement algorithm could enhance and dehaze the low-illumination video of coal mine heading face in real time, which avoids the problems of dimness, distortion, blurring and mutation of video images. Compared with Retinex algorithm, ALTM algorithm and dark channel prior algorithm, the proposed video enhancement algorithm significantly improves the information entropy, standard deviation and average gradient of the video image, and has a higher real-time processing speed, which can provide high-quality and reliable support for subsequent processings such as video target recognition, target tracking, target monitoring and image segmentation of heading face video.