A Visibility Measurement Method Based on Automatic Image Recognition

Abstract

The invention discloses a visibility measurement method based on automatic image recognition, which collects an original image containing a target and a test image; uses the maximum class variance method to perform threshold segmentation on the original image and the test image; and uses a diamond probe to perform threshold segmentation on the test image. Continuous expansion to obtain an expanded image; inverse coloring of the expanded image, using the principle of centroid, to find the "pseudo" centroid calibration of the multi-connected regions after continuous expansion; take the "pseudo" centroid as the center of the multi-target region to be selected, and mark it for selection In the dark channel rendering of the original image after segmentation, find the center point with the largest "pseudo" centroid light intensity in the corresponding multi-target area marked to be selected, and the area where the center point is located is used for visibility measurement. In the final target area, select the appropriate dehazing coefficient ω to calculate the visibility. It can select the most suitable defogging coefficient for the images of different scenes captured by the camera equipment, and measure the visibility.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to a visibility measurement method based on automatic image recognition. Background technique [0002] With the continuous advancement of the reform and opening up process, the rapid economic development and the increase of urban population, the energy consumption has increased rapidly, resulting in aggravation of air pollution and reduced visibility, which has brought serious problems to people's daily life and other aspects. Influence, especially in aerospace, transportation and other aspects of great harm. Although various measures have been taken to control it nationwide, it is quite difficult. In recent years, it can still be seen that haze weather occurs frequently in various places. [0003] There are three commonly used visibility detection methods: manual visual inspection, instrument measurement and digital camera method. The visual measurement metho...

AI Technical Summary

Problems solved by technology

The visual method has large errors and is greatly affected by the personal circumstances of the observer; the instrumental method is expensive and has high site requirements, and cannot be widely used; the digital camera method is practical, low in cost, and can overcome the influence of bad weather, so it is favored by experts at home and abroad. Scholars study extensively

Images