改進型DSSD算法在道路損傷檢測中的應用研究
2021年電子技術應用第12期
蘇 可1,郭學俊2,楊 瑩3,陳澤華2
1.太原理工大學 電氣與動力工程學院,山西 太原030024; 2.太原理工大學 大數據學院,山西 晉中030600;3.山西省交通科技研發有限公司,山西 太原030006
摘要: 在自動檢測中,由于道路損傷數據集存在小目標損傷難檢測與類別不平衡問題,導致道路損傷檢測的準確率低、虛假率高。為此,在DSSD(Deconvolutional Single Shot Detector)網絡模型的基礎上,提出一種結合注意力機制和Focal loss的道路損傷檢測算法。首先,采用識別精度更高的ResNet-101作為DSSD模型的基礎網絡;其次,在ResNet-101主干網絡中添加注意力機制,采用通道域注意力和空間域注意力結合的方式,實現特征在通道維度上的加權與空間維度上的聚焦,提升對小目標道路損傷的檢測效果;最后,為了減少簡單樣本的權重,增大難分類樣本的權重,使用Focal loss來提高整體的檢測效果。在Global Road Damage Detection Challenge比賽所提供的數據集上進行驗證,實驗結果表明,該模型的平均精度均值為83.95%,比基于SSD和YOLO網絡的道路損傷檢測方法的準確率更高。
中圖分類號: TP391.41
文獻標識碼: A
DOI:10.16157/j.issn.0258-7998.211684
中文引用格式: 蘇可,郭學俊,楊瑩,等. 改進型DSSD算法在道路損傷檢測中的應用研究[J].電子技術應用,2021,47(12):64-68,99.
英文引用格式: Su Ke,Guo Xuejun,Yang Ying,et al. Research on application of improved DSSD algorithm in road damage detection[J]. Application of Electronic Technique,2021,47(12):64-68,99.
文獻標識碼: A
DOI:10.16157/j.issn.0258-7998.211684
中文引用格式: 蘇可,郭學俊,楊瑩,等. 改進型DSSD算法在道路損傷檢測中的應用研究[J].電子技術應用,2021,47(12):64-68,99.
英文引用格式: Su Ke,Guo Xuejun,Yang Ying,et al. Research on application of improved DSSD algorithm in road damage detection[J]. Application of Electronic Technique,2021,47(12):64-68,99.
Research on application of improved DSSD algorithm in road damage detection
Su Ke1,Guo Xuejun2,Yang Ying3,Chen Zehua2
1.College of Electrical and Power Engineering,Taiyuan University of Technology,Taiyuan 030024,China; 2.College of Data Science,Taiyuan University of Technology,Jinzhong 030600,China; 3.Shanxi Transportation Technology Research and Development Co.,Ltd.,Taiyuan 030006,China
Abstract: In the automatic detection, the road damage data set has the problems of difficult detection of small target damage and imbalance of categories, resulting in low accuracy and high false rate of road damage detection. For this reason, based on the DSSD(deconvolutional single shot detector) network model, a road damage detection algorithm combining attention mechanism and Focal loss is proposed. First of all, ResNet-101 with higher recognition accuracy is used as the basic network of the DSSD model. Secondly, an attention mechanism is added to the ResNet-101 backbone network, and the channel domain attention and spatial domain attention are combined to achieve the weighting of features in the channel dimension and the focus on the spatial dimension, and improve the detection effect of small target road damage. Finally, in order to reduce the weight of simple samples and increase the weight of difficult-to-classify samples, Focal loss is used to improve the overall detection effect. It is verified on the data set provided by the Global Road Damage Detection Challenge competition. The experimental results show that the average accuracy of the model is 83.95%, which is more accurate than the road damage detection method based on SSD and YOLO network.
Key words : road damage detection;DSSD target detection algorithm;small target detection;attention mechanism;category imbalance problem
0 引言
道路建設是衡量國家現代化水平的重要指標之一,我國道路交通網龐大復雜,道路養護問題凸顯。如何從道路圖像快速準確地檢測出損傷區域及類型成為學者研究的熱點。
隨著深度學習的快速發展,使用卷積神經網絡(Convolution Neural Network,CNN)[1]自主地從數據集中提取相應特征信息成為主流方法,如快速的R-CNN[2]、SSD[3]、YOLO[4]等。這些網絡能夠定位和識別圖中具有邊界框的對象,為復雜背景下道路檢測提供了有效的框架。
本文詳細內容請下載:http://www.viuna.cn/resource/share/2000003873。
作者信息:
蘇 可1,郭學俊2,楊 瑩3,陳澤華2
(1.太原理工大學 電氣與動力工程學院,山西 太原030024;
2.太原理工大學 大數據學院,山西 晉中030600;3.山西省交通科技研發有限公司,山西 太原030006)
此內容為AET網站原創,未經授權禁止轉載。