Open Access

Research on Intelligent Recognition System of Traffic Image Based on CNN and Intelligent Recognition of Foreign Object Intrusion

Jinlin Tan

Jinlin Tan

Shaanxi Aerospace Technology Application Research Institute Co., LtdXi'an, China
Search for this author on
Sciendo|Google Scholar
Tan, Jinlin
, 
Liang Wang

Liang Wang

Shaanxi Aerospace Technology Application Research Institute Co., LtdXi'an, China
Search for this author on
Sciendo|Google Scholar
Wang, Liang
, 
Xiaotian Yang

Xiaotian Yang

Shaanxi Aerospace Technology Application Research Institute Co., LtdXi'an, China
Search for this author on
Sciendo|Google Scholar
Yang, Xiaotian
, 
Yunfei Song

Yunfei Song

Shaanxi Aerospace Technology Application Research Institute Co., LtdXi'an, China
Search for this author on
Sciendo|Google Scholar
Song, Yunfei
, 
Weiming Wang

Weiming Wang

Shaanxi Aerospace Technology Application Research Institute Co., LtdXi'an, China
Search for this author on
Sciendo|Google Scholar
Wang, Weiming
 and 
Xin Yu

Xin Yu

Shaanxi Aerospace Technology Application Research Institute Co., LtdXi'an, China
Search for this author on
Sciendo|Google Scholar
Yu, Xin
  
Mar 19, 2025
Research on Intelligent Recognition System of Traffic Image Based on CNN and Intelligent Recognition of Foreign Object Intrusion's Cover Image
About this article
Previous Article
Next Article
Cite
Download Cover
Published Online: Mar 19, 2025
Received: Oct 31, 2024
Accepted: Feb 06, 2025
DOI: https://doi.org/10.2478/amns-2025-0400
Keywords
© 2025 Jinlin Tan et al., published by Sciendo
This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Figure 1.

Overall technical route
Overall technical route

Figure 2.

The calculation process of the coordinate of the track pixels
The calculation process of the coordinate of the track pixels

Figure 3.

Calculation process of sensitive area
Calculation process of sensitive area

Figure 4.

The structure of the convolution neural network
The structure of the convolution neural network

Figure 5.

Validation of different classic models on data sets
Validation of different classic models on data sets

Figure 6.

Identification of different algorithms in the test set
Identification of different algorithms in the test set

Figure 7.

The exact rate of foreign matter recognition in different methods
The exact rate of foreign matter recognition in different methods

Comparison of foreign body judgment accuracy

Methods fr fz Pr
Light flow method 1265 1865 67.83%
Three frame difference method 1318 1865 70.67%
KNN background difference method 1591 1865 85.31%
Hollow filling KNN background difference method 1645 1865 88.20%
Our method 1722 1865 92.33%

Identification results of different foreign invasion

Algorithms Faster R-CNN SSD300 YOLOv5s Our method
mAP (%) 80.55 81.49 83.28 88.47
Model size(M) 619.81 192.93 98.12 72.24
Test time(ms) 287.52 87.58 49.85 26.01
Airplane 76.13 70.35 78.99 84.55
Bicycle 81.72 70.14 78.51 85.95
Bird 84.14 87.96 80.87 86.78
Boat 82.69 80.72 90.22 93.46
Bottle 80.07 86.95 77.57 89.42
Bus 78.26 76.59 88.21 90.05
Car 86.55 87.49 89.77 92.67
Cat 72.91 81.89 82.73 85.14
Chair 81.45 78.58 80.49 84.43
Cow 83.31 80.47 82.26 85.34
Table 83.96 82.02 83.77 87.92
Dog 79.08 83.81 90.12 92.65
Horse 81.77 84.12 81.63 87.63
Motorcycle 86.59 83.24 82.37 88.81
Pedestrians 88.58 78.68 80.41 94.65
Potted plant 74.75 78.47 80.66 84.97
Sheep 83.93 92.64 87.32 90.81
Sofa 74.11 83.62 80.39 86.82
Train 77.84 80.59 86.74 90.44
TV 73.19 81.56 82.52 86.84
Language:
English
Publication timeframe:
1 times per year
Journal Subjects:
Life Sciences, Life Sciences, other, Mathematics, Applied Mathematics, General Mathematics, Physics, Physics, other
Journal RSS Feed