Open AccessResearch On Grooved Rail Garbage Identification Algorithm Based On Improved YOLOv3
Author(s) -
Xiaojie Huang,
Kangquan Ye,
Zhongbin Fang,
Yongjun Xie,
Xizhe Ma,
Jing Ji,
Qiantong Wu
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1827/1/012185
Subject(s) - garbage , computer science , identification (biology) , robustness (evolution) , cluster analysis , block (permutation group theory) , artificial intelligence , algorithm , pattern recognition (psychology) , computer vision , mathematics , biochemistry , chemistry , botany , geometry , biology , gene , programming language
To solve the problem that existing modern tram track cleaning vehicles cannot automatically and accurately identify grooved rail garbage, this paper proposes a grooved rail garbage identification algorithm based on improved YOLOv3. This algorithm firstly extracts the groove rail region in the image, then adjusts and optimizes the residual block and convolutional layer network in the basic network of YOLOv3, and adopts k-means algorithm for clustering analysis to obtain anchor value of the adaptive dataset, which effectively improved the average identification accuracy of grooved rail garbage. The experimental results showed that the improved YOLOv3 model mAP reached 89.16% on the self-made Grooved Rail Garbage Dataset. For the image of 416×416, the recognition speed reached more than 14 frames per second. Compared with the traditional grooved rail garbage identification algorithm, this algorithm not only has higher recognition rate and accuracy, but also has good robustness to environmental changes.