Vehicle target recognition algorithm for UAV image based on DRFP

被引：0

作者：

Zhang Z. ^{[1
,2
,3
,4
,5
]}

Liu Y. ^{[1
,2
,4
,5
]}

Wang S. ^{[1
,2
,3
,4
,5
]}

Liu T. ^{[1
,2
,3
,4
,5
]}

Lin Z. ^{[1
,2
,3
,4
,5
]}

机构：

[1] Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang

[2] Institutes for Robotics and Intelligent Manufacturing, Chinese Academy of Sciences, Shenyang

[3] University of Chinese Academy of Sciences, Beijing

[4] Key Laboratory of Opto-electronic Information Processing, Chinese Academy Sciences, Shenyang

[5] The Key Lab of Image Understanding and Computer Vision, Shenyang

来源：

Hongwai yu Jiguang Gongcheng/Infrared and Laser Engineering | 2019年 / 48卷

关键词：

Deep learning; Small target recognition; UAV-image;

D O I：

10.3788/IRLA201948.S226001

中图分类号：

学科分类号：

摘要：

In order to solve the problem of small target recognition caused by small size, less edge and texture information in the field of view for UAV in complex battlefield environment, a new model based on deep learning for small target recognition Deep Residual and Feature Pyramid (DRFP) was proposed in this paper. Firstly, the residual structure was used as the skeleton of the model, and the feature pyramid structure was used to achieve feature fusion. Secondly, the cross-entropy function with adjusting factor was used in the loss function to realize the focus of attention on difficult samples. Finally, a non-maximum Gaussian suppression algorithm was used to improve the detection rate of target-intensive areas. The experimental results show that the accuracy (mAP) of proposed single stage model is 83.16% using UAV-images towards vehicle recognition, which achieves the level of two stage network model. At the same time, the recognition speed meets real-time requirements. © 2019, Editorial Board of Journal of Infrared and Laser Engineering. All right reserved.

引用

共 10 条

[1] Yao W., Liu Y., Zhu C., Deep learning of full-reference image quality assessment based on human visual properties, Infrared and Laser Engineering, 47, 7, (2018)
[2] Yu Y., Kong L., Zhang H., Et al., An improved material removal model for robot polishing based on neural networks, Infrared and Laser Engineering, 48, 3, (2019)
[3] Geng L., Liang X., Xiao Z., Et al., Real-time driver fatigue detection based on morphology infrared features and deep learning, Infrared and Laser Engineering, 47, 2, (2018)
[4] Girshick R., Donahue J., Darrell T., Et al., Rich Feature Hierarchies for Accurate Object Detection and Semantic segmentation, IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 1, pp. 580-587, (2014)
[5] Ren S., He K., Girshick R., Et al., Faster R-CNN: Towards real-time object detection with region proposal networks, IEEE Transactions on Pattern Analysis & Machine Intelligence, 39, 6, pp. 1137-1149, (2015)
[6] Redmon J., Farhadi A., YOLOv3: An incremental improvement
[7] Terrail J.O.D., Jurie, Frederic, Faster RER-CNN: application to the detection of vehicles in aerial images
[8] He K., Zhang X., Ren S., Et al., Deep Residual Learning for Image recognition, IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 1, pp. 770-778, (2015)
[9] Sandler M., Howard A., Zhu M., Et al., MobileNetV2: Inverted Residuals and Linear bottlenecks, IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 1, pp. 4510-4520, (2018)
[10] Zhang X., Zhou X., Lin M., Et al., ShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile devices, IEEE Conference on Computer Vision and Pattern Recognition(CVPR), 1, pp. 6848-6856, (2018)

← 1 →