Open AccessDetection method of dense bridge disease targets based on SE-YOLOv3
Author(s) -
Yaolin Wang,
Zhang Zhi,
Hongcheng Yin
Publication year - 2020
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 - 10.1088/1742-6596/1544/1/012141
Subject(s) - computer science , artificial intelligence , pyramid (geometry) , bridge (graph theory) , pattern recognition (psychology) , set (abstract data type) , process (computing) , feature (linguistics) , data set , computer vision , mathematics , medicine , linguistics , philosophy , geometry , programming language , operating system
In this paper, we find that the overlapped detection frames of dense disease in the bridge disease data set, combined with the current research on target detection algorithms in dense scene. Based on YOLOv3, the feature pyramid is optimized to manually set thresholds for post-processing after extracting features. A bridge disease target detection model based on SE-YOLOv3 was proposed. In the post-processing process, IoU uses Soft-IoU to detect the accuracy of the prediction box, approves the value of IoU, and then uses the EM algorithm based on the Gaussian distribution to perform a function operation to delete the low-score prediction detection box to achieve the purpose of removing overlap. Experimental results show that the detection accuracy is 95.73%.