Improved Infrared Anomaly Target Detection Algorithm Based on Single Gaussian Model
-
摘要: 基于单高斯模型的红外异常目标检测算法是一种常见的能自适应更新背景模型的检测算法。该算法对各个像素的输出响应进行高斯建模,通过设定的阈值确定目标像素点是否为前景像素点,从而达到检测的目的。本文在单高斯模型的基础上,提出一种改进的异常检测算法,该算法利用奈曼-皮尔逊准则选取最佳阈值,克服了根据经验值选取阈值的局限性,为最佳判决阈值的选取奠定了理论基础,使得在虚假率一定的情况下,检测概率达到最高。实验证明,将常见的经验阈值与本文确定阈值进行比较,本文算法确定的阈值检测效果更佳。Abstract: An infrared anomaly target detection algorithm based on a single Gaussian model is a commonly used detection algorithm that can adaptively update the background model. The algorithm performs Gaussian modeling on the output response of each pixel and determines whether the target pixel is a foreground pixel through a defined threshold to realize detection. This paper proposes an improved anomaly detection algorithm based on a single Gaussian model. The algorithm uses the Neiman-Pearson criterion to define the optimal threshold, which overcomes the limitation of selecting the threshold based on empirical values. The paper lays a theoretical foundation for obtaining the best decision threshold so that under a certain false rate, the detection probability can reach the highest value. Experimental results show that, compared to the commonly experienced thresholds, the threshold determined in this study provides a much better detection effect.
-
Key words:
- infrared imaging /
- target detection /
- single Gaussian model /
- Neiman-Pearson criterion
-
表 1 二元信号检测判决结果
Table 1. Judgment results of binary signal detection
Judgment Assumption H0 H1 H0 (H0/H0) (H0/H1) H1 (H1/H0) (H1/H1) 表 2 不同阈值的检测判决结果
Table 2. Detection and judgment results of different thresholds
Threshold TP TN FP FN A MA FA Threshold1 93 3259 1766 2 0.6547 0.0211 0.9499 Threshold2 88 4511 514 7 0.8982 0.0737 0.8538 Threshold3 75 4959 80 20 0.9805 0.2105 0.5161 -
[1] 庞文正. 红外热成像在民用领域应用不断拓展市场前景不可限量[J]. 中国安防, 2020, 15(3): 62-65. https://www.cnki.com.cn/Article/CJFDTOTAL-ZGAF202003013.htmPANG Wenzheng. The application of infrared thermal imaging in the civil field continues to expand and the market prospects are unlimited[J]. China Security, 2020, 15(3): 62-65. https://www.cnki.com.cn/Article/CJFDTOTAL-ZGAF202003013.htm [2] 刘成, 鲍可进. 基于图像处理的嵌入式自动报警系统[J]. 计算机工程与设计, 2007, 28(17): 4198-4199, 4269. doi: 10.3969/j.issn.1000-7024.2007.17.045LIU Cheng, BAO Kejin. Embedded automatic alarm system based on image processing[J]. Computer Engineering and Design, 2007, 28(17): 4198-4199, 4269. doi: 10.3969/j.issn.1000-7024.2007.17.045 [3] 谈新权, 江柳, 刘伟宏. 红外热成像报警方法研究[J]. 激光与红外, 1997, 27(5): 286-291. https://www.cnki.com.cn/Article/CJFDTOTAL-JGHW199705012.htmTAN Xinquan, JIANG Liu, LIU Weihong. Research on infrared thermal imaging alarm method[J]. Laser and Infrared, 1997, 27(5): 286-291. https://www.cnki.com.cn/Article/CJFDTOTAL-JGHW199705012.htm [4] 许彬, 郑链, 王永学, 等. 红外序列图像小目标检测与跟踪技术综述[J]. 红外与激光工程, 2004, 33(5): 44-49. https://www.cnki.com.cn/Article/CJFDTOTAL-HWYJ200405011.htmXU Bin, ZHENG Lian, WANG Yongxue, et al. Overview of small target detection and tracking technology in infrared serial images[J]. Infrared and Laser Engineering, 2004, 33(5): 44-49. https://www.cnki.com.cn/Article/CJFDTOTAL-HWYJ200405011.htm [5] LI H, WEI Y T, LI L Q, et al. Infrared moving target detection and tracking based on tensor locality preserving projection[J]. Infrared Physics & Technology, 2010, 53(2): 77-83. http://www.onacademic.com/detail/journal_1000034076425010_63d8.html [6] Sengar S S, Mukhopadhyay S. Moving object detection using statistical background subtraction in wavelet compressed domain[J]. Multimedia Tools & Applications, 2019, 12(12): 1-22. doi: 10.1007/s11042-019-08506-z?utm_medium=affiliate&utm_content=meta [7] 张晓露, 李玲, 辛云宏. 基于小波变换的自适应多模红外小目标检测[J]. 激光与红外, 2017, 47(5): 647-652. doi: 10.3969/j.issn.1001-5078.2017.05.023ZHANG Xiaolu, LI Ling, XIN Yunhong. Adaptive multi-mode infrared small target detection based on wavelet transform[J]. Laser and Infrared, 2017, 47(5): 647-652. doi: 10.3969/j.issn.1001-5078.2017.05.023 [8] 秦剑, 陈钱, 钱惟贤. 基于光流估计和自适应背景抑制的弱小目标检测[J]. 光子学报, 2011, 40(3): 476-482. https://www.cnki.com.cn/Article/CJFDTOTAL-GZXB201103032.htmQIN Jian, CHEN Qian, QIAN Weixian. Dim target detection based on optical flow estimation and adaptive background suppression[J]. Acta Photonica Sinica, 2011, 40(3): 476-482. https://www.cnki.com.cn/Article/CJFDTOTAL-GZXB201103032.htm [9] 詹令明, 李翠芸, 姬红兵. 基于显著图的红外弱小目标动态规划检测前跟踪算法[J]. 计算机辅助设计与图形学学报, 2019, 31(7): 1061-1066. https://www.cnki.com.cn/Article/CJFDTOTAL-JSJF201907001.htmZHAN Lingming, LI Cuiyun, JI Hongbing. Infrared dim target tracking algorithm based on saliency map dynamic programming before detection[J]. Journal of Computer Aided Design and Graphics, 2019, 31(7): 1061-1066. https://www.cnki.com.cn/Article/CJFDTOTAL-JSJF201907001.htm [10] 刘源, 李庆, 梁艳菊. 基于FPGA的红外目标自动检测系统[J]. 红外技术, 2019, 41(6): 521-526. http://hwjs.nvir.cn/article/id/hwjs201906005LIU Yuan, LI Qing, LIANG Yanju. Infrared target automatic detection system based on FPGA[J]. Infrared Technology, 2019, 41(6): 521-526. http://hwjs.nvir.cn/article/id/hwjs201906005 [11] 徐鹏, 任波. 基于单高斯模型的森林火灾烟图像目标检测[J]. 计算机与现代化, 2009(2): 18-20. doi: 10.3969/j.issn.1006-2475.2009.02.006XU Peng, REN Bo. Forest fire smoke image target detection based on single Gaussian model[J]. Computer and Modernization, 2009(2): 18-20. doi: 10.3969/j.issn.1006-2475.2009.02.006 [12] 朱华生, 叶军. 基于改进高斯模型的车流量检测算法[J]. 激光与红外, 2013, 43(10): 1180-1183. doi: 10.3969/j.issn.1001-5078.2013.10.22ZHU Huasheng, YE Jun. Vehicle flow detection algorithm based on improved Gaussian model[J]. Laser & Infrared, 2013, 43(10): 1180-1183. doi: 10.3969/j.issn.1001-5078.2013.10.22 [13] 喻旭勇, 王直杰. 一种基于改进单高斯模型的运动目标提取方法[J]. 科学技术与工程, 2013, 13(13): 3609-3613. doi: 10.3969/j.issn.1671-1815.2013.13.017YU Xuyong, WANG Zhijie. A moving target extraction method based on improved single Gaussian model[J]. Science Technology and Engineering, 2013, 13(13): 3609-3613. doi: 10.3969/j.issn.1671-1815.2013.13.017 [14] 王小平, 张丽杰, 常佶. 基于单高斯背景模型运动目标检测方法的改进[J]. 计算机工程与应用, 2009, 45(21): 122-124. https://www.cnki.com.cn/Article/CJFDTOTAL-JSGG200921035.htmWANG Xiaoping, ZHANG Lijie, CHANG Ji. Improvement of moving target detection method based on single Gaussian background model[J]. Computer Engineering and Applications, 2009, 45(21): 122-124. https://www.cnki.com.cn/Article/CJFDTOTAL-JSGG200921035.htm [15] 吴一全, 孟天亮, 吴诗婳. 图像阈值分割方法研究进展20年(1994—2014)[J]. 数据采集与处理, 2015, 30(1): 1-23. https://www.cnki.com.cn/Article/CJFDTOTAL-SJCJ201501001.htmWU Yiquan, MENG Tianliang, WU Shihua. Research progress of image threshold segmentation methods for 20 years (1994-2014)[J]. Data Acquisition and Processing, 2015, 30(1): 1-23. https://www.cnki.com.cn/Article/CJFDTOTAL-SJCJ201501001.htm [16] 李文杰, 闫世强, 胡磊, 等. 红外预警卫星系统虚警抑制技术综述[J]. 红外技术, 2020, 42(2): 115-120. http://hwjs.nvir.cn/article/id/hwjs202002002LI Wenjie, YAN Shiqiang, Hu Lei, et al. Overview of false alarm suppression technology of infrared early warning satellite system[J]. Infrared Technology, 2020, 42(2): 115-120. http://hwjs.nvir.cn/article/id/hwjs202002002 [17] 鞠德航, 林可祥, 陈捷. 信号检测理论导论[M]. 北京: 科学出版社, 1997.JU Dehang, LIN Kexiang, CHEN Jie. Introduction to Signal Detection Theory[M]. Beijing: Science Press, 1997.