|
A Traffic Light Recognition Method Based on Prior Information
ZENG Yangfan
2022, 47(21):
37-42.
DOI: 10.16638/j.cnki.1671-7988.2022.021.007
To solve the problem of low accuracy of color-based traffic light recognition (TLR), a
TLR method that combines HSV (Hue, Saturation, Value) color space and gray image information
is proposed. First, image tilt correction and size standardization is needed to meet the needs of the
proposed fusion traffic light recognition method. Then, the RGB (Red, Green, Blue) images are
converted into HSV (Hue, Saturation, Value) color space to initially recognize the state of traffic
light. The boundaries of traffic lights is extracted by hough line detection, and region of interest (ROI) detection is realized. Considering that some traffic light images cannot be recognized by
HSV (Hue, Saturation, Value) color space method, and the structure characteristics of the traffic
light are used as the prior information, and the region of interest pixel values of the gray image is
summed to further identify the state of traffic light. The fusion traffic light recognition method is
evaluated through experiments and compared with the methods that separately use HSV (Hue,
Saturation, Value) color space and gray image region of interest pixel values summation to evaluate
the performance of the proposed algorithm.
References |
Related Articles |
Metrics
|