Segmentation Method of Forbidden Traffic Signs Based on MSPCNN Model with Adjustable Parameters

Jing Di; Jinghui Wang; Jing Lian

doi:10.3788/LOP202158.0215002

Laser & Optoelectronics Progress, Volume. 58, Issue 2, 0215002(2021)

Segmentation Method of Forbidden Traffic Signs Based on MSPCNN Model with Adjustable Parameters

Jing Di, Jinghui Wang^*, and Jing Lian

Author Affiliations

School of Electronic and Information Engineering, Lanzhou Jiaotong University, Lanzhou, Gansu 730070, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(6)

Fig. 1. Reddening preprocessing effect. (a) Original images; (b) corresponding reddening preprocessing effect

Download full size

Fig. 2. Experimental results of different methods under uniform illumination. (a) Original images; (b) PA-MSPCNN; (c) OTSU; (d) SPCNN; (e) PCNN

Download full size

Fig. 3. Experimental results of different methods under uneven illumination. (a) Original images; (b) PA-MSPCNN; (c) OTSU; (d) SPCNN; (e) PCNN

Download full size

Table 1. Evaluation metrics of four different images obtained by different methods under uniform illumination

View table

Table 1. Evaluation metrics of four different images obtained by different methods under uniform illumination

Image name	Method	Recall rate	Precision rate	F-measure	Jaccard coefficient
3317	PA-MSPCNN	0.9942	0.2221	0.3631	0.2218
	OTSU	0.7855	0.2073	0.3281	0.1962
	SPCNN	0.7702	0.2068	0.3260	0.1948
	PCNN	0.9898	0.2182	0.3575	0.2177
5917	PA-MSPCNN	0.9043	0.9994	0.9495	0.9039
	OTSU	0.7622	0.9965	0.8637	0.7601
	SPCNN	0.7563	0.9976	0.8604	0.7549
	PCNN	0.9052	0.9858	0.9438	0.8936
42577	PA-MSPCNN	0.9980	0.5018	0.6678	0.5013
	OTSU	0.8275	0.9910	0.9019	0.8213
	SPCNN	0.9672	0.3890	0.5548	0.3839
	PCNN	0.9840	0.2674	0.4206	0.2663
90579	PA-MSPCNN	0.7282	0.8575	0.7876	0.6496
	OTSU	0.5607	0.9110	0.6941	0.5316
	SPCNN	0.5335	0.8848	0.6656	0.4988
	PCNN	0.5335	0.8917	0.6676	0.5010

Table 2. Evaluation metrics of four different images obtained by different methods under uneven illumination

View table

Table 2. Evaluation metrics of four different images obtained by different methods under uneven illumination

Image name	Method	Recall rate	Precision rate	F-measure	Jaccard coefficient
6889	PA-MSPCNN	0.9980	0.4949	0.6617	0.4944
	OTSU	0.0039	0.0351	0.0071	0.0036
	SPCNN	0.2934	0.5091	0.3723	0.2287
	PCNN	0.2934	0.5091	0.3723	0.2287
15434	PA-MSPCNN	0.8259	0.7232	0.7712	0.6276
	OTSU	0.3024	0.8371	0.4443	0.2856
	SPCNN	0.2997	0.8314	0.4406	0.2825
	PCNN	0.2989	0.8311	0.4396	0.2817
Image name	Method	Recall rate	Precision rate	F-measure	Jaccard coefficient
33884	PA-MSPCNN	0.9424	0.8358	0.8859	0.7952
	OTSU	0.0996	0.8581	0.1785	0.0980
	SPCNN	0.0917	0.8592	0.1657	0.0903
	PCNN	0.0917	0.8592	0.1657	0.0903
45424	PA-MSPCNN	0.9889	0.7671	0.8640	0.7606
	OTSU	0.1807	0.9974	0.3060	0.1806
	SPCNN	0.1898	0.8514	0.3104	0.1837
	PCNN	0.1856	0.9928	0.3127	0.1854

Table 3. Recognition accuracy of different methods
View table
Table 3. Recognition accuracy of different methods
Method Recognition accuracy/%
OTSU 63
SPCNN 61
PCNN 60
PA-MSPCNN 85

Tools

Get Citation

Copy Citation Text

Jing Di, Jinghui Wang, Jing Lian. Segmentation Method of Forbidden Traffic Signs Based on MSPCNN Model with Adjustable Parameters[J]. Laser & Optoelectronics Progress, 2021, 58(2): 0215002

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Machine Vision

Received: Mar. 23, 2020

Accepted: May. 8, 2020

Published Online: Jan. 11, 2021

The Author Email: Wang Jinghui (455342316@qq.com)

DOI:10.3788/LOP202158.0215002

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Evaluation metrics of four different images obtained by different methods under uniform illumination

Table 1. Evaluation metrics of four different images obtained by different methods under uniform illumination

Table 2. Evaluation metrics of four different images obtained by different methods under uneven illumination

Table 2. Evaluation metrics of four different images obtained by different methods under uneven illumination

Table 3. Recognition accuracy of different methods

Table 3. Recognition accuracy of different methods