Template matching monitoring video defogging system and its defogging method based on cloud platform

Abstract

Disclosed are a template matching surveillance video defogging system and method based on a cloud platform. Defogging is conducted in combination with an adaptive clustering K-means algorithm and a dark channel prior algorithm, thereby reducing the performance demands and power consumption of a terminal device. The present invention achieves the beneficial effects as follows: when key frames are defogged, the key frames are divided on the basis of fog concentration and defogged independently, thereby weakening the influence on the profile, shape and size of a background object, and weakening the mutual influence between adjacent sceneries having different depths in a defogging process; when non-key frames are defogged, the invariance of a viewing angle and the spatial-temporal correlativity of each frame are fully utilized, the computation speed is greatly increased by training a special transmittance template library of each terminal device, thereby making it very effective to defog a foggy video acquired at a fixed viewing angle; and the power loss and performance requirements of the terminal devices are reduced, and a good platform is provided for achieving a better defogging effect and processing a larger number of video data.

Description

Technical field [0001] The invention relates to a template matching monitoring video defogging system and a defogging method based on a cloud platform, and belongs to the technical field of video processing. Background technique [0002] With the development of image processing and computer vision technology, the demand for outdoor vision systems represented by road traffic video surveillance systems has increased; haze weather has largely limited the role of video surveillance systems; current video defogging systems are rare Considering the inter-frame correlation and its image structure information, the process of processing has a strong dependence on the collection of weather and scene information, and is complex to the environment (such as occlusion, excessive gap in the depth of the scene in the video, and the sky area in the video) The defogging of the video is not good; the defogging operation of the video puts forward higher requirements on the performance of the device,...

AI Technical Summary

Problems solved by technology

[0002] With the development of image processing and computer vision technology, the demand for outdoor vision systems represented by road traffic video surveillance systems has increased; fog and haze weather has largely limited the role of video surveillance systems; there are few current video defogging systems Considering its inter-frame correlation and its image structure information, the processing process has a strong dependence on the collection of weather and scene information, and the environment is complex (such as occlusion, too large depth gap of the scene in the video, and the sky area in the video) The defogging operation of the video is not good; the defogging operation of the video puts forward higher requirements on the performance of the equipment, and it is necessary to find an effective video defogging method and system

[0003] At present, the research on video defogging methods has achieved many meaningful results. For example, the Langley Research Center in the United States has conducted in-depth research on the domain-based Retinex algorithm to enhance foggy videos. However, it does not take into account the temporal and spatial correlation of video image sequences and frames. Inner image structure features, there are redundant processing and dependence on the scene environment, and greatly consume the performance and power of the terminal device

Images