Real-Time Traffic Sign Detection Method Based on Improved Convolution Neural Network

Ying Tong; Huicheng Yang

doi:10.3788/LOP56.071003

Journals >Laser & Optoelectronics Progress >Volume 56 >Issue 7 >Page 071003 > Article

Laser & Optoelectronics Progress
Vol. 56, Issue 7, 071003 (2019)

Real-Time Traffic Sign Detection Method Based on Improved Convolution Neural Network

Ying Tong^* and Huicheng Yang

Author Affiliations

College of Electrical Engineering, Anhui Polytechnic University, Wuhu, Anhui 241000, China

show less

DOI: 10.3788/LOP56.071003 Cite this Article Set citation alerts

Ying Tong, Huicheng Yang. Real-Time Traffic Sign Detection Method Based on Improved Convolution Neural Network[J]. Laser & Optoelectronics Progress, 2019, 56(7): 071003 Copy Citation Text

show less

Fig. 1. Flow chart of program

Download full size

Fig. 2. Structural diagram of NF R-CNN neural network

Download full size

Fig. 3. Three examples of traffic signs. (a) Mandatory; (b) prohibitory; (c) danger

Download full size

Fig. 4. Examples of test results by three models. (a) ZF; (b) VGG16; (c) NF R-CNN

Download full size

Fig. 5. P-R curves of three examples. (a) Mandatory; (b) prohibitory; (c) danger

Download full size

Item	Threshold Θ
Item	0.1	0.2	0.4	0.5	0.6	0.7
Precision	0.7654	0.8826	0.9542	0.9688	09835	1
Recall	0.9542	0.9403	0.9125	0.8623	0.7854	0.6281

Table 1. Accuracy and recall rates at different time thresholds

Model		Mandatory /%	Prohibitory /%	Danger /%	Time /s
Model-ZF	P	91.89	97.68	98.65	0.133
Model-ZF	R	79.11	70.56		93.15
Model-VGG16	P	98.85	99.44	99.86	0.130
Model-VGG16	R	75.23	69.26		87.84
Model-NF R-CNN	P	98.10	99.22	99.23	0.016
Model-NF R-CNN	R	80.31	72.96		94.78

Table 2. Comparison of classification results by three models

Reference	AUC value /%			Time/s
Reference		Mandatory	Prohibitory	Danger
Ref. [8]	96.54	99.35	96.68	0.15
Ref. [9]	92.56	96.58	91.27	0.31
Ref. [10]	93.47	100	98.64	0.66
Ref. [11]	100	100	99.22	3.47
NF R-CNN	97.02	99.55	99.37	0.016

Table 3. AUC values and processing time

Ying Tong, Huicheng Yang. Real-Time Traffic Sign Detection Method Based on Improved Convolution Neural Network[J]. Laser & Optoelectronics Progress, 2019, 56(7): 071003

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information