Low-visibility recognition algorithm based on live-action image

Abstract

The invention discloses a low-visibility recognition algorithm based on a live-action image, and relates to the technical field of low-visibility recognition algorithms of live-action images, the low-visibility recognition algorithm comprises a visual management system, and the visual management system further comprises an algorithm for collecting a live-action image based on video camera equipment and extracting visibility information from the live-action image for recognition. The method comprises the following steps: S1, starting: eliminating interference information such as characters of a live-action image, and performing gradient conversion on the live-action image to obtain an image gradient map; S2, partitioning: horizontally partitioning the image from top to bottom, wherein each block has the same size. The algorithm provided by the invention has general applicability, and most real images can be distinguished by using one model; the learning efficiency is high, a new simplified image is established for learning by eliminating redundant information of an original image, and the learning training efficiency is improved; and the system also has a good recognition capability for live-action images acquired by night videos, and is suitable for popularization and application.

Description

technical field [0001] The invention relates to the technical field of low-visibility recognition algorithms for real-scene images, in particular to a low-visibility recognition algorithm based on real-scene images. Background technique [0002] Visibility is the visible distance of the target, which refers to the maximum distance at which the outline of the target can be distinguished from the background when the target is observed. Visibility is one of the meteorological observation items. Low visibility has many adverse effects on transportation such as ferries, civil aviation, highways, power supply, and even the daily life of citizens. In today's highly developed economy, the impact of low visibility weather is more obvious. In the past 10 years, major traffic accidents caused by low visibility have occurred frequently. Therefore, accurate monitoring and acquisition of visibility data are of great significance to traffic and shipping. The existing automatic observat...

AI Technical Summary

Problems solved by technology

[0005] (1) It is aimed at visibility recognition in a specific scene, such as video image visibility recognition on a highway, and the recognition algorithm is not universal. For different scenes, it is necessary to obtain samples again to learn and build different models;

[0006] (2) The ability to recognize visibility at night is poor, and the ability to recognize image visibility in night light environments is relatively weak

Images