Low illumination model identification method based on Retinex and S-SIFT characteristic combination

Abstract

The invention relates to the computer vision field and particularly relates to a model identification method under the low illumination environment based on Retinex and S-SIFT characteristic combination. The method comprises four steps that 1), low illumination image enhancement based on Retinex and wavelet adaptive threshold transformation is carried out; 2), license plate positioning is carried out; 3), interception of a vehicle face region including the model information is carried out; and 4), an S-SIFT characteristic model of model region images is established, and model identification is carried out in combination with an SVM trainer. The method is advantaged in that an embedded FPGA can be utilized for realization, the method is applied to a camera or a video camera which has a model identification function to output images in real time under the low illumination environment, system accuracy and reliability are effectively improved, and real-time demands can be satisfied.

Description

technical field [0001] The invention relates to the field of computer vision, in particular to a method for identifying vehicle types under low illumination conditions based on the combination of Retinex and S-SIFT (sparse scaleinvariant feature transform) features. Background technique [0002] With the rapid development of modern transportation, security and other industries, people pay more and more attention to automatic vehicle identification technology, which is one of the important research topics of computer vision and pattern recognition technology in the field of intelligent transportation in recent years. The automatic identification system of vehicle models can be used for vehicle management in highway toll stations, parking lots, crossroads and other places, and can also be used for vehicle entry and exit management in modern residential areas or industrial parks. It is very important for public safety, community security, road traffic and parking lot vehicle man...

AI Technical Summary

Problems solved by technology

[0004] In reality, the pictures collected in practice usually have complex backgrounds, uneven lighting, low resolution, old vehicles, dirty vehicles, etc., especially at night or when the weather is dark, due to low light illumination, the vehicle video images are blurred, At the same time, video surveillance systems generally have different angles for vehicle image capture, which brings great difficulties to vehicle identification.

Images