Dim Small Object Detection Method Based on Statistical Feature Space Extraction and SVM
-
摘要: 小天体检测是小天体防御和预警的前提。针对小天体目标信噪比低、检测难的问题，提出了基于统计特征空间提取和支持向量机（SVM）的极暗弱小天体检测方法。区别于传统方法基于时间或空间上目标的能量和背景噪声能量的瞬时能量差别或是瞬时能量差别的累积，对目标进行检测。该方法不依赖目标能量大小，提取运动目标穿过背景时对稳定性产生的扰动来反演运动目标。将输入的图像序列转化为单像元时序信号，划分时序窗口提取统计特征，关联形成统计特征空间，利用更高维度的变化特性检测目标变化。通过SVM将暗弱小天体检测问题转化为目标与背景的二分类问题，避开了较难解决的阈值分割问题同时具有更好的泛化性能。利用真实数据与其他经典方法进行对比分析，使得分类准确率提高4.02%。该方法能够适应更低的信噪比，在极低信噪比下仍表现出稳定的检测性能。Abstract: Small object detection is the premise of small object defense and early warning. Aiming at the problems of low signal-to-noise ratio and difficult detection of small object targets, a very dark and weak small object detection method based on statistical feature space extraction and SVM is proposed. It is different from traditional methods to detect targets based on the instantaneous energy difference or the accumulation of instantaneous energy difference between target energy and background noise energy in time or space. This method does not directly depend on the target energy, but extracts the disturbance to the stability when the moving target passes through the background to retrieve the moving target. The input image sequence is transformed into a single pixel timing signal, the timing window is divided, the statistical features are extracted, the statistical feature space is formed by correlation, and the target change is detected by using the change characteristics of higher dimensions. The method is implemented by Support Vector Machine (SVM) transforms the dim small object detection problem into the binary classification problem of target and background, so that the method avoids the thorny threshold segmentation problem and has better generalization performance. Through the comparative analysis between real data and other classical methods, the classification accuracy is increased by 4.02%. This method can adapt to lower signal-to-noise ratio and still perform well at very low signal-to-noise ratio Stable detection performance.
-
图 1 目标背景的不同特征分布
Figure 1. Distribution of different characteristics of target background
表 1 不同方法实验结果对比
Table 1. Comparison of experimental results of different methods
方法 虚警率/(%) 检测率/(%) AUC 背景建模 16.30 84.61 0.9193 Top-hat滤波 8.54 92.80 0.9771 最大中值滤波 22.50 86.10 0.9357 最大均值滤波 16.45 88.71 0.9431 小波变换 9.68 91.17 0.9677 文中方法 3.23 96.82 0.9915 注 AUC指ROC曲线与（1，0）和（1，1）点围绕的面积。 
下载: 导出CSV
表 2 不同方法实验结果对比
Table 2. Comparison of experimental results of different methods
方法 信噪比 正样本分类
准确率/(%)负样本分类
准确率/(%)背景建模 37.14 63.16 96.97 Top-hat滤波 29.28 42.08 97.46 最大均值滤波 61.95 53.77 99.47 本文方法 98.44 74.28 99.84
下载: 导出CSV
-
[1] XUE D N, SUN J Q, HU Y Q, et al. Dim small target detection based on convolutinal neural network in star image[J]. Multimedia Tools and Applications, 2020, 79(7): 4681-4698 [2] LI Y S, LI Z Z, ZHANG C, et al. Infrared maritime dim small target detection based on spatiotemporal cues and directional morphological filtering[J]. Infrared Physics & Technology, 2021, 115: 103657 [3] BAI X Z, ZHOU F G. Analysis of new top-hat transformation and the application for infrared dim small target detection[J]. Pattern Recognition, 2010, 43(6): 2145-2156 doi: 10.1016/j.patcog.2009.12.023 [4] MENG W, JIN T, ZHAO X W. Adaptive method of dim small object detection with heavy clutter[J]. Applied Optics, 2013, 52(10): D64-D74 doi: 10.1364/AO.52.000D64 [5] XU Y L, WANG W H. A method for single frame detection of infrared dim small target in complex background[J]. Journal of Physics: Conference Series, 2020, 1634(1): 012063 doi: 10.1088/1742-6596/1634/1/012063 [6] QI S X, MA J, TAO C, et al. A robust directional saliency-based method for infrared small-target detection under various complex backgrounds[J]. IEEE Geoscience and Remote Sensing Letters, 2013, 10(3): 495-499 doi: 10.1109/LGRS.2012.2211094 [7] YANG C C, MA J Y, ZHANG M F, et al. Multiscale facet model for infrared small target detection[J]. Infrared Physics & Technology, 2014, 67: 202-209 [8] LI H, TAN Y H, LI Y S, et al. Image layering based small infrared target detection method[J]. Electronics Letters, 2014, 50(1): 42-44 doi: 10.1049/el.2013.3042 [9] CHEN C L P, LI H, WEI Y T, et al. A local contrast method for small infrared target detection[J]. IEEE Transactions on Geoscience and Remote Sensing, 2014, 52(1): 574-581 doi: 10.1109/TGRS.2013.2242477 [10] GAO C Q, MENG D Y, YANG Y, et al. Infrared patch-image model for small target detection in a single image[J]. IEEE Transactions on Image Processing, 2013, 22(12): 4996-5009 doi: 10.1109/TIP.2013.2281420 [11] HAN J H, MA Y, ZHOU B, et al. A robust infrared small target detection algorithm based on human visual system[J]. IEEE Geoscience and Remote Sensing Letters, 2014, 11(12): 2168-2172 doi: 10.1109/LGRS.2014.2323236 [12] HE L H, GE L. CamShift target tracking based on the combination of inter-frame difference and background difference[C]//2018 37 th Chinese Control Conference (CCC). Wuhan, China: IEEE, 2018: 889-893 [13] CHEN H, BOHUSH R P, CHEN C, et al. Estimation of people movement in video based on optical flow block method and motion maps[J]. Pattern Recognition and Image Analysis, 2021, 31(2): 261-270 doi: 10.1134/S105466182102005X [14] ZHANG T, SONG C F. Object detection based on minimum convex hull and generalized Hough transform[J]. IOP Conference Series: Materials Science and Engineering, 2020, 768(7): 072058 doi: 10.1088/1757-899X/768/7/072058 [15] TESTOLIN A, DIAMANT R. Combining denoising autoencoders and dynamic programming for acoustic detection and tracking of underwater moving targets)[J]. Sensors, 2020, 20(10): 2945 doi: 10.3390/s20102945 [16] BOSQUET B, MUCIENTES M, BREA V M. STDnet-ST: Spatio-temporal ConvNet for small object detection[J]. Pattern Recognition, 2021, 116: 107929 doi: 10.1016/j.patcog.2021.107929 [17] DU J M, LI D Y, DENG Y J, et al. Multiple frames based infrared small target detection method using CNN[C]//2021 4 th International Conference on Algorithms, Computing and Artificial Intelligence (ACAi 2021). Sanya, China: Association for Computing Machinery, 2021: 397-402 [18] KUMAR S S, ABRAHAM D M, JAHANSHAHI M R, et al. Automated defect classification in sewer closed circuit television inspections using deep convolutional neural networks[J]. Automation in Construction, 2018, 91: 273-283 doi: 10.1016/j.autcon.2018.03.028 [19] REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: Unified, real-time object detection[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Las Vegas, NV, USA: IEEE, 2015: 779-788 [20] REDMON J, FARHADI A. YOLO9000: better, faster, stronger[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Honolulu, HI, USA: IEEE, 2016: 6517-6525 [21] LIU W, ANGUELOV D, ERHAN D, et al. SSD: Single shot MultiBox detector[C]//14 th European Conference on Computer Vision. Amsterdam, The Netherlands: Springer, 2016: 21-37 [22] GIRSHICK R. Fast R-CNN[C]//2015 IEEE International Conference on Computer Vision. Santiago, Chile: IEEE, 2015: 1440-1448 [23] REN SQ, HE K M, GIRSHICK R, et al. Faster R-CNN: Towards real-time object detection with region proposal networks[C]//International Conference on Neural Information Processing Systems. Cambridge: MIT Press, 2015: 91-99 [24] HE K M, GEORGIA G, DOLLÁR P, et al. Mask R-CNN[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42(2): 386-397 doi: 10.1109/TPAMI.2018.2844175 [25] PATAN G, BARSKY B A. An Introduction to Laplacian Spectral Distances and Kernels: Theory, Computation, and Applications[M]. San Rafael: Morgan & Claypool Publishers, 2017 [26] DESHPANDE S D, ER M H, VENKATESWARLU R, et al. Max-mean and max-median filters for detection of small targets[J]. Proceedings of the SPIE, 1999, 3809: 74-83 doi: 10.1117/12.364049 [27] XU Y, LUO M Z, LI T, et al. ECG signal De-noising and baseline wander correction based on CEEMDAN and Wavelet Threshold[J]. Sensors (Basel, Switzerland), 2017, 17(12): 2754 doi: 10.3390/s17122754 [28] BOUWMANS T, EL BAF F, VACHON B. Background modeling using mixture of Gaussians for foreground detection-a survey[J]. Recent Patents on Computer Science, 2008, 1(3): 219-237 doi: 10.2174/2213275910801030219 [29] HADHOUD M M, THOMAS D W. The two-dimensional adaptive LMS (TDLMS) algorithm[J]. IEEE Transactions on Circuits and Systems, 1988, 35(5): 485-494 doi: 10.1109/31.1775 [30] SONI T, ZEIDLER J R, KU W H. Performance evaluation of 2-D adaptive prediction filters for detection of small objects in image data[J]. IEEE Transactions on Image Processing, 1993, 2(3): 327-340 doi: 10.1109/83.236534 -
下载:
点击查看大图
图(7) / 表(2)
计量
- 文章访问数: 156
- HTML全文浏览量: 54
- PDF下载量: 36
- 被引次数: 0
