Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

< Back to Article

Fig 1.

The proposed algorithm.

More »

Fig 1 Expand

Fig 2.

ROI setting result based on lane detection algorithm.

(a) Transformation method. (b) Result of applying the lane detection algorithm. (c) Result of masking operation based on the derived ROI.

More »

Fig 2 Expand

Fig 3.

Result of color space conversion and thresholding.

(a) Original image. (b) a* channel after converting the original image into L*a*b* color space. (c) Result of applying the gamma correction with γ = 10 to the original a* channel. (d) Result of applying threshold method directly to the a* channel of the image. (e) Result of applying threshold method to a* channel to which gamma correction is applied.

More »

Fig 3 Expand

Fig 4.

a* channel histogram and threshold value when the Otsu threshold method is applied.

(a) When no preprocessing is applied. (b) When gamma correction is applied.

More »

Fig 4 Expand

Fig 5.

Gamma correction mapping curves.

(a) Gamma correction mapping curve when 0 < γ ≤ 1. (b) Gamma correction mapping curve when 1 ≤ γ.

More »

Fig 5 Expand

Fig 6.

Symmetry test result.

(a) Result of applying the masking operation and the morphological operation to the binary image to which thresholding is applied. (b) Result of drawing bounding boxes based on the remaining pixels. (c) Result of finding the optimal rear-lights pair among the candidates through the symmetry test. (d) The final result of rear-lights region detection.

More »

Fig 6 Expand

Fig 7.

Calculation of pixel-level similarity between two candidate regions using correlation coefficients.

More »

Fig 7 Expand

Fig 8.

Rear-Lights region detection results and feature extraction of detected rear-lights regions.

From the top, the first row is a grayscale image of rear-lights regions, and the second row is the rear-lights regions image with HSV color range filtering applied. From the left, the first column is brake on condition at daytime, the second column is brake off condition at daytime, the third column is brake on condition at nighttime, and the fourth column is brake off condition at nighttime.

More »

Fig 8 Expand

Fig 9.

Comparing the mean of HSV channels under brake on and off conditions.

(a) Mean of each HSV color channel under brake on and off conditions. (b) Maximum density point for each distribution of brake on and off conditions.

More »

Fig 9 Expand

Fig 10.

Rear-lights regions image with HSV color range filtering applied when turn signals are on.

More »

Fig 10 Expand

Fig 11.

Sum of mean S channel value and V channel value in the brake on and off conditions.

More »

Fig 11 Expand

Table 1.

Rear-lights region detection algorithm evaluation result.

More »

Table 1 Expand

Fig 12.

Rear-lights region detection failure cases.

(a) Failed to detect rear-lights region due to the distance from the vehicle in front. (b) Failed to detect rear-lights region due to raindrops.

More »

Fig 12 Expand

Table 2.

Comparative evaluation result of rear-lights region detection algorithm.

More »

Table 2 Expand

Fig 13.

Comparative evaluation result of rear-lights region detection algorithm.

More »

Fig 13 Expand

Fig 14.

Relationship between recall score and precision score.

(a) Recall score and precision score were evaluated while changing the threshold value. (b) Precision-recall curve (PR curve).

More »

Fig 14 Expand

Fig 15.

Brake-lights detection result.

(a) Brake on condition in a day environment. (b) Brake off condition in a day environment. (c) Brake on condition in a night environment. (d) Brake off condition in a night environment. (e) Brake on condition in a cloudy environment. (f) Brake off condition in a cloudy environment. (g) Brake on situation in a rainy environment. (h) Brake off condition in a rainy environment.

More »

Fig 15 Expand

Table 3.

Brake-lights detection algorithm evaluation result.

More »

Table 3 Expand

Table 4.

Comparison of processing speed.

More »

Table 4 Expand

Table 5.

Proportion of time required by algorithms.

More »

Table 5 Expand