Intelligent Detection of "Problematic Map" Using Convolutional Neural Network

被引：0
作者：
Ren J. [1 ,2 ,3 ]
Liu W. [1 ]
Li Z. [2 ]
Li R. [1 ]
Zhai X. [1 ]
机构：
[1] National Geomatics Center of China, Beijing
[2] Faculty of Geosciences and Environmental Engineering, Southwest Jiaotong University, Chengdu
[3] School of Geosciences and Info-Physics, Central South University, Changsha
来源：
Liu, Wanzeng (luwnzg@163.com) | 1600年 / Editorial Board of Medical Journal of Wuhan University卷 / 46期
关键词：
problematic map; Convolutional neural network (CNN); Multi-scale feature fusion; Object detection; Small sample scene;
D O I：
10.13203/j.whugis20190259
中图分类号：
学科分类号：
摘要：
Objectives: In order to solve the problem of low efficiency of manual visual discrimination in the audit of "problematic map" in China, an end-to-end adaptive detection method for small sample scene based on the multi-scale feature fusion of convolutional neural network (CNN) is proposed in this paper. Methods: The real-time enhancement of the dataset can overcome the shortcoming of CNN, which requires a large number of training samples. By fusing multiple maps at different scales, the intelligent detection of significant error areas of the "problematic map" in multiple scales is realized. The region proposal network is optimized by taking the attributes of wrong regions into account to further improve the detection accuracy. Results: Compared with the existing detection method, the accuracy of the proposed method is increased by 8 times, which verifies its effectiveness. Conclusions: The proposed method provides a new solution for large-scale "problematic map" detection, and can be quickly applied in production. © 2021, Editorial Board of Geomatics and Information Science of Wuhan University. All right reserved.
引用
收藏
页码：570 / 577
页数：7
相关论文
共 19 条
[1]  
Zhou Jiawei, Wu Xiaochun, Current Situation Analysis and Countermeasure of "Problem Map, Geomatics Technology and Equipment, 20, 4, pp. 42-43, (2018)
[2]  
Huang Long, Research on the Method of Fast Mining Internet Problem Map Picture, Geomatics & Spatial Information Technology, 40, 11, pp. 92-93, (2017)
[3]  
Krizhevsky A, Sutskever I, Hinton G E., ImageNet Classification with Deep Convolutional Neural Networks, Communications of the ACM, 60, 6, pp. 84-90, (2017)
[4]  
Szegedy C, Liu W, Jia Y, Et al., Going Deeper with Convolutions, IEEE Conference on Computer Vision and Pattern Recognition, (2015)
[5]  
He K, Zhang X, Ren S, Et al., Deep Residual Learning for Image Recognition, IEEE Conference on Computer Vision and Pattern Recognition, (2016)
[6]  
Huang G, Liu Z, Weinberger K Q, Et al., Densely Connected Convolutional Networks, IEEE Conference on Computer Vision and Pattern Recognition, (2017)
[7]  
Hu J, Shen L., Squeeze-and-Excitation Networks, IEEE Conference on Computer Vision and Pattern Recognition, (2018)
[8]  
Zhang L L, Lin L, Liang X D, Et al., Is Faster R-CNN Doing Well for Pedestrian Detection?, Computer Vision - ECCV 2016, pp. 443-457, (2016)
[9]  
Li Qinghui, Li Aihua, Wang Tao, Et al., Double-Stream Convolutional Networks with Sequential Optical Flow Image for Action Recognition, Acta Optica Sinica, 38, 6, pp. 226-232, (2018)
[10]  
Schmidhuber J., Deep Learning in Neural Networks: An Overview, Neural Networks, 61, pp. 85-117, (2015)
12