Implementation of a Modified Faster R-CNN for Target Detection Technology of Coastal Defense Radar
Abstract
:1. Introduction
- (1)
- The deep learning method is applied to the measured data of the coastal defense radar with a low resolution, and its detection results are compared with the two–dimensional CFAR detectors.
- (2)
- ResNet50 is replaced with a relatively shallow backbone feature extraction network VGG16 in the Faster R-CNN, and a parametric rectified linear unit (PReLU) function is used to achieve more specialized activations.
- (3)
- (4)
- The Soft-NMS [22] algorithm is used to eliminate the possible overlapped detection boxes.
2. Methods
2.1. CFAR Detector
2.1.1. CA-CFAR
2.1.2. GOCA-CFAR and SOCA-CFAR
2.2. Target Detection Based on FASTER R-CNN
2.2.1. ResNet50 Feature Extraction Network
2.2.2. Faster R-CNN
2.3. Modified Faster R-CNN Structure
2.3.1. VGG16 Feature Extraction Network
2.3.2. K-Means Clustering Algorithm Initializes Anchors
2.3.3. Soft-NMS
Algorithm 1. Soft-NMS | |
Input: The list of initial detection boxes , and the corresponding detection scores | |
1: | Set: D = |
2: | whiledo |
3: | |
4: | |
5: | |
6: | for in B do |
7: | |
8: | end for |
9: | end while |
Output:D, S |
3. Experiments
3.1. Target Detection Steps in a Coastal Defense Radar Image
- (1)
- Preprocess the radar echo signals and transform the echo data into a pulse-range two-dimensional (2D) image for subsequent network training and testing.
- (2)
- Make the training set label and the test set division. For the pulse-range 2D echo data image, the echo image is segmented according to the pulse number and under the condition that the target does not cross the range unit, and the aspect ratio of the image is close. They are cooperative targets; thus, their bounding boxes are determined according to the target position information provided by the GPS. We then use the annotation software, LabelImg, to complete the labeling work on the segmented amplitude image.
- (3)
- Initialize the basic network parameters and use the K-means clustering algorithm to set the anchor size.
- (4)
- Start the CNN training and use the gradient descent algorithm to calculate the error between the output of the network and the real target, then backpropagate the error to adjust the network parameters, such as the weight and the bias. Continue the training until the network converges or reaches the preset training times.
- (5)
- Verify whether or not the network is underfitting or overfitting during the training process using the verification set of each epoch and adjusting the network parameters to continue training.
- (6)
- After training, obtain the target detector of the coastal defense radar based on the Faster R-CNN. The Soft-NMS algorithm is added to eliminate the overlapping detection boxes and test the sample set. The target is detected from the radar amplitude image. We calculate the recognition accuracy rate, false-alarm rate, recall rate, and mean average precision (mAP) to determine whether or not the evaluation index is met. If it is satisfied, the target detector has completed the training and can be used for the target detection in the unknown sea-surface radar image; otherwise, return to step 3 to adjust the network parameters for retraining.
3.2. Build Data Set
3.3. Experimental Environment and Parameters
3.4. Evaluation Criteria and Results
3.5. Further Comparative Experiments
4. Discussion
5. Conclusions
Author Contributions
Funding
Conflicts of Interest
References
- Sperle, M.; Negri, E.; Ternes, C. Automatic classification of sidescan sonar images for mapping marine mineral resources. In Proceedings of the 2015 IEEE/OES Acoustics in Underwater Geosciences Symposium (RIO Acoustics), Rio de Janeiro, Brazil, 29–31 July 2015; pp. 1–5. [Google Scholar]
- von Jouanne, A.; Brekken, T.K.A. Ocean and geothermal energy systems. Proc. IEEE 2017, 105, 2147–2165. [Google Scholar] [CrossRef]
- Ward, J.; Schultz, I.; Woodruff, D.; Roesijadi, G.; Copping, A. Assessing the effects of marine and hydrokinetic energy development on marine and estuarine resources. In Proceedings of the OCEANS 2010 MTS/IEEE SEATTLE, Seattle, WA, USA, 20–23 September 2010; pp. 1–7. [Google Scholar]
- Yu, X.; Chen, X.; Huang, Y.; Guan, J. Fast Detection Method for Low-Observable Maneuvering Target via Robust Sparse Fractional Fourier Transform. IEEE Geosci. Remote Sens. Lett. 2020, 17, 978–982. [Google Scholar] [CrossRef]
- Xu, S.W.; Shi, X.Y.; Shui, P.L. An adaptive detector with mismatched signals rejection in compound Gaussian clutter. J. Radars 2019, 8, 326–334. [Google Scholar]
- Chen, X.; Guan, J.; Huang, Y.; Xue, Y.; Liu, N. Radar signal processing for low-observable marine target-challenges and solutions. In Proceedings of the 2019 IEEE International Conference on Signal, Information and Data Processing (ICSIDP), Chongqing, China, 11–13 December 2019; pp. 1–6. [Google Scholar]
- Wang, Z.; Liu, S. A new smooth method for the l1 exact penalty function for inequality constrained optimization. In Proceedings of the 2010 Third International Joint Conference on Computational Science and Optimization, Huangshan, Anhui, China, 28–31 May 2010; pp. 110–113. [Google Scholar]
- Simonyan, K.; Zisserman, A. Very deep convolutional networks for large-scale image recognition. arXiv 2014, arXiv:1409.1556. [Google Scholar]
- Ning, C.; Zhou, H.; Song, Y.; Tang, J. Inception Single Shot MultiBox Detector for object detection. In Proceedings of the 2017 IEEE International Conference on Multimedia & Expo Workshops (ICMEW), Hong Kong, China, 10–14 July 2017; pp. 549–554. [Google Scholar]
- He, K.M.; Zhang, X.Y.; Ren, S.Q.; Sun, J. Deep residual learning for image recognition. In Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA, 27–30 June 2016; pp. 770–778. [Google Scholar]
- Girshick, R.; Donahue, J.; Darrell, T.; Malik, J. Rich feature hierarchies for accurate object detection and semantic segmentation. In Proceedings of the 2014 IEEE Conference on Computer Vision and Pattern Recognition, Columbus, OH, USA, 23–28 June 2014; pp. 580–587. [Google Scholar]
- Girshick, R. Fast R-CNN. In Proceedings of the 2015 IEEE International Conference on Computer Vision (ICCV), Santiago, Chile, 7–13 December 2015; pp. 1440–1448. [Google Scholar]
- Ren, S.Q.; He, K.M.; Girshick, R.; Sun, J. Faster R-CNN: Towards real-time object detection with region proposal networks. IEEE Trans. Pattern Anal. Mach. Intell. 2017, 39, 1137–1149. [Google Scholar] [CrossRef] [PubMed] [Green Version]
- Loi, N.V.; Trung Kien, T.; Hop, T.V.; Thanh Son, L.; Khuong, N.V. Abnormal Moving Speed Detection Using Combination of Kernel Density Estimator and DBSCAN for Coastal Surveillance Radars. In Proceedings of the 2020 7th International Conference on Signal Processing and Integrated Networks (SPIN), Noida, India, 27–28 February 2020; pp. 143–147. [Google Scholar]
- Weiss, M. Analysis of some modified cell-averaging CFAR processors in multiple-target situations. IEEE Trans. Aerosp. Electron. Syst. 1982, 18, 102–114. [Google Scholar] [CrossRef]
- Matthias, K.; Hermann, R. Fast two-dimensional CFAR procedure. IEEE Trans. Aerosp. Electron. Syst. 2013, 6, 1817–1823. [Google Scholar]
- Yan, J.W.; Li, X.; Shao, Z.H. Intelligent and fast two-dimensional CFAR procedure. In Proceedings of the 2015 IEEE International Conference on Communication Problem-Solving (ICCP), Guilin, China, 16–18 October 2015; pp. 461–463. [Google Scholar]
- Mou, X.; Chen, X.; Guan, J.; Chen, B.; Dong, Y. Marine target detection based on improved faster R-CNN for navigation radar PPI images. In Proceedings of the 2019 International Conference on Control, Automation and Information Sciences (ICCAIS), Chengdu, China, 23–26 October 2019; pp. 1–5. [Google Scholar]
- Zhu, D.Y.; Xiang, T.S.; Wei, W.; Ren, Z.W.; Yang, M.D.; Zhu, Z.D. An extended two step approach to high-resolution airborne and spaceborne SAR full-aperture processing. IEEE Trans. Geosci. Remote Sens. 2020, 99, 1–16. [Google Scholar] [CrossRef]
- Sasikumar, P.; Khara, S. K-Means clustering in wireless sensor networks. In Proceedings of the 2012 Fourth International Conference on Computational Intelligence and Communication Networks, Mathura, India, 3–5 November 2012; pp. 140–144. [Google Scholar]
- Zhong, Y.Y.; Wang, J.F.; Peng, J.; Zhang, L. Anchor Box Optimization for Object Detection. In Proceedings of the 2020 IEEE Winter Conference on Applications of Computer Vision (WACV), Snowmass, CO, USA, 1–5 March 2020; pp. 1275–1283. [Google Scholar]
- Bodla, N.; Singh, B.; Chellappa, R.; Davis, L.S. Soft-NMS—Improving Object Detection with One Line of Code. In Proceedings of the 2017 IEEE International Conference on Computer Vision (ICCV), Venice, Italy, 22–29 October 2017; pp. 5562–5570. [Google Scholar]
- Wang, G.; Giannakis, G.B.; Chen, J. Learning ReLU Networks on Linearly Separable Data: Algorithm, Optimality, and Generalization. IEEE Trans. Signal Process. 2019, 67, 2357–2370. [Google Scholar] [CrossRef] [Green Version]
- Alex, K.; Sutskever, I.; Hinton, G.E. ImageNet Classification with Deep Convolutional Neural Networks. Int. Conf. Neural Inf. Process. Syst. 2012, 25, 1097–1105. [Google Scholar]
- He, K.M.; Zhang, X.Y.; Ren, S.Q.; Sun, J. Delving deep into rectifiers: Surpassing human-level performance on imagenet classification. In Proceedings of the 2015 IEEE International Conference on Computer Vision (ICCV), Santiago, Chile, 7–13 December 2015; pp. 1026–1034. [Google Scholar]
- Yann, L.; Bengio, Y.; Hinton, G.E. Deep learning. Nature 2015, 521, 436–444. [Google Scholar]
- Neubeck, A.; Van Gool, L. Efficient non-maximum suppression. In Proceedings of the 18th International Conference on Pattern Recognition (ICPR’06), Hong Kong, China, 20–24 August 2006; pp. 850–855. [Google Scholar]
- Jae, H.S.; Kyung, R.O.; In, K.K.; Jang, Y.L. Application of maritime AIS (Automatic Identification System) to ADS-B (Automatic Dependent Surveillance-Broadcast) transceiver. In Proceedings of the 2010 International Conference on Control, Automation, and Systems (ICCAS), Gyeonggido, Korea, 27–30 October 2010; pp. 2233–2237. [Google Scholar]
System | Windows 10 | Tool | Anaconda 3 |
---|---|---|---|
RAM | 32 GB | Programming | Python 3.6 |
CPU | Intel i5-9600KF @3.7 GHz × 6 | IDE | VS code |
GPU | NVIDIA GTX 1080Ti 11 G | Framework | Pytorch-GPU |
Auxiliary tools | MATLAB | Others | CUDA 10.0 |
Detection Algorithm | /% | /% | T/s | ||||
---|---|---|---|---|---|---|---|
Sample 1 | CA-CFAR | 9 | 255 | 10 | 3.40 | 90 | 0.9569 |
GOCA-CFAR | 9 | 207 | 10 | 4.16 | 90 | 1.3561 | |
SOCA-CFAR | 9 | 699 | 10 | 1.27 | 90 | 1.3853 | |
Sample 2 | CA-CFAR | 4 | 155 | 4 | 2.51 | 100 | 0.9594 |
GOCA-CFAR | 4 | 129 | 4 | 3.00 | 100 | 1.3583 | |
SOCA-CFAR | 4 | 339 | 4 | 1.16 | 100 | 1.3347 | |
Sample 3 | CA-CFAR | 12 | 162 | 13 | 6.89 | 92.30 | 0.9752 |
GOCA-CFAR | 12 | 143 | 13 | 7.74 | 92.30 | 1.3653 | |
SOCA-CFAR | 13 | 385 | 13 | 3.26 | 100 | 1.3520 | |
Sample 4 | CA-CFAR | 9 | 121 | 9 | 6.92 | 100 | 0.9582 |
GOCA-CFAR | 9 | 103 | 9 | 8.03 | 100 | 1.5329 | |
SOCA-CFAR | 9 | 181 | 9 | 4.73 | 100 | 1.5749 | |
Sample 5 | CA-CFAR | 1 | 97 | 2 | 1.02 | 50 | 0.9541 |
GOCA-CFAR | 1 | 83 | 2 | 1.19 | 50 | 1.3584 | |
SOCA-CFAR | 2 | 99 | 2 | 1.98 | 100 | 1.3374 | |
Sample 6 | CA-CFAR | 3 | 114 | 3 | 2.56 | 100 | 0.9534 |
GOCA-CFAR | 3 | 99 | 3 | 2.94 | 100 | 1.3754 | |
SOCA-CFAR | 3 | 116 | 3 | 2.52 | 100 | 1.3626 |
T/s | ||||||
---|---|---|---|---|---|---|
Sample 1 | 9 | 0 | 10 | 100 | 90 | 0.6562 |
Sample 2 | 4 | 0 | 4 | 100 | 100 | 0.5983 |
Sample 3 | 12 | 0 | 13 | 100 | 92.31 | 0.6661 |
Sample 4 | 9 | 0 | 9 | 100 | 100 | 0.7041 |
Sample 5 | 2 | 0 | 2 | 100 | 100 | 0.6692 |
Sample 6 | 3 | 0 | 3 | 100 | 100 | 0.5734 |
Method | T/s | mAP/% | |||||
---|---|---|---|---|---|---|---|
ResNet50 | 342 | 55 | 394 | 86.15 | 86.80 | 0.7042 | 81.41 |
ResNet50 + K-means | 367 | 42 | 394 | 89.73 | 93.15 | 0.8317 | 87.20 |
VGG16 | 320 | 22 | 394 | 93.57 | 81.22 | 0.2648 | 81.52 |
VGG16 + K-means | 370 | 35 | 394 | 91.36 | 93.91 | 0.4559 | 91.42 |
Improved Faster R-CNN | 370 | 32 | 394 | 92.04 | 93.91 | 0.3904 | 92.27 |
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |
© 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Yan, H.; Chen, C.; Jin, G.; Zhang, J.; Wang, X.; Zhu, D. Implementation of a Modified Faster R-CNN for Target Detection Technology of Coastal Defense Radar. Remote Sens. 2021, 13, 1703. https://doi.org/10.3390/rs13091703
Yan H, Chen C, Jin G, Zhang J, Wang X, Zhu D. Implementation of a Modified Faster R-CNN for Target Detection Technology of Coastal Defense Radar. Remote Sensing. 2021; 13(9):1703. https://doi.org/10.3390/rs13091703
Chicago/Turabian StyleYan, He, Chao Chen, Guodong Jin, Jindong Zhang, Xudong Wang, and Daiyin Zhu. 2021. "Implementation of a Modified Faster R-CNN for Target Detection Technology of Coastal Defense Radar" Remote Sensing 13, no. 9: 1703. https://doi.org/10.3390/rs13091703
APA StyleYan, H., Chen, C., Jin, G., Zhang, J., Wang, X., & Zhu, D. (2021). Implementation of a Modified Faster R-CNN for Target Detection Technology of Coastal Defense Radar. Remote Sensing, 13(9), 1703. https://doi.org/10.3390/rs13091703