基于局部对比度机制的红外弱小目标检测算法

韩金辉; 董兴浩; 蒋亚伟; 李知铮; 梁琨; 张利红

基于局部对比度机制的红外弱小目标检测算法

1.
周口师范学院物理与电信工程学院，河南周口 466001
2.
华中科技大学电子信息与通信学院，湖北武汉 430074

基金项目:

国家自然科学基金 61802455

国家自然科学基金 62003381

河南省科技厅科技发展计划项目 192102210089

河南省教育厅高等学校重点科研项目 18B510021

周口师范学院大学生创新创业训练计划项目 S202010478010

详细信息

作者简介:
韩金辉（1986-），男，河南周口人，讲师，博士，主要从事红外小目标检测、红外图像处理等研究。E-mail：hanjinhui@zknu.edu.cn

中图分类号: TP391
计量
- 文章访问数: 523
- HTML全文浏览量: 131
- PDF下载量: 105
- 被引次数: 0
出版历程
- 收稿日期: 2020-04-07
- 修回日期: 2021-04-06
- 刊出日期: 2021-04-20

Infrared Small Dim Target Detection Based on Local Contrast Mechanism

1.
College of Physics and Telecommunication Engineering, Zhoukou Normal University, Zhoukou 466001, China
2.
School of Electronic Information and Communications, Huazhong University of Science and Technology, Wuhan 430074, China

摘要

摘要: 针对复杂背景和低信杂比条件下的红外弱小目标检测难题，提出了一种基于局部对比度机制的红外弱小目标检测方法。该方法提出了一个包含中心层、中间层和最外层的3层窗口，可以使用单尺度计算完成不同尺度弱小目标的检测。首先，对中心层引入匹配滤波思想，有针对性地增强真实目标；同时，提出最接近滤波原则，对最外层进行背景估计，以缓解目标靠近边缘时的检测难题；然后，在目标增强结果与背景估计结果之间进行比差联合的对比度计算，达到同时增强目标和抑制背景的目的；最后，通过自适应阈值分割，提取真实目标。实验结果表明，相比现有算法而言，该算法可更好地增强目标、抑制复杂背景，且原理简洁易实现，可有效减少运算量。
- 红外小目标 /
- 目标检测 /
- 局部对比度 /
- 目标增强 /
- 背景抑制
Abstract: A method for infrared (IR) small dim target detection based on a local contrast mechanism is proposed to solve the problem of IR small dim target detection under a complex background and low signal-to-clutter ratio (SCR). A three-layer window consisting of an inner layer, a middle layer, and an outer layer is proposed, so that targets of different scales can be detected using only single-scale calculations. First, the matched filter is applied to the inner layer to enhance the true target purposefully, and the max-close principle is proposed to estimate the background of the outer layer, so that detection becomes easier when the target is near the background edge. Then, the ratio-difference joint local contrast measure is calculated between the enhanced target and the estimated background to enhance the true target and suppress the complex background simultaneously. Finally, an adaptive threshold operation is used to extract the true target. Experimental results show that compared to some existing algorithms, the proposed algorithm can enhance the true target and suppress complex background better, and its principle is simple yet suitable for implementation and can effectively reduce the amount of calculation.
- IR small target /
- target detection /
- local contrast /
- target enhancement /
- background suppression

HTML全文

图 1 本文算法框图

Figure 1. The flowchart of the proposed algorithm

下载: 全尺寸图片幻灯片

图 2 三层窗口

Figure 2. Three-layer window

下载: 全尺寸图片幻灯片

图 3 匹配滤波器模板

Figure 3. Template of the matched filter

下载: 全尺寸图片幻灯片

图 4 窗口处于不同位置的示意图，灰色区域代表普通背景，白色区域代表高亮背景

Figure 4. Different cases when the window at different positions, the grey area represents the normal background, the white area represents the bright background

下载: 全尺寸图片幻灯片

图 5 本文算法各个阶段处理后的图像（序列）

Figure 5. The processing result of each stage using the proposed algorithm(sequence)

下载: 全尺寸图片幻灯片

图 6 本文算法各个阶段处理后的图像（单帧数据库）

Figure 6. The processing result of each stage using the proposed algorithm(single-frame dataset)

下载: 全尺寸图片幻灯片

图 7 本文算法对于特殊情况检测失败的示例

Figure 7. Example of fail detection for the proposed algorithm under special condition

下载: 全尺寸图片幻灯片

图 8 不同序列的ROC曲线：(a)~(j)：序列1~序列10

Figure 8. ROC curves of different sequences: (a)-(j): Seq. 1-Seq. 10

下载: 全尺寸图片幻灯片

表 1 10组红外序列的详细信息

Table 1. Details of the ten IR sequences

Seq	Frames	Size	Target size	Target type
1	200	320×240	7×5-4×3	Plane
2	200	320×256	3×3	Plane
3	200	320×256	3×5	Plane
4	70	256×200	4×5-4×10	Plane
5	200	256×320	2×3-3×4	Plane
6	300	256×256	3×3-3×4	Unmanned vehicle
7	300	256×320	2×2-3×3	Vehicle
8	100	256×256	3×5	Vehicle
9	100	256×256	5×5-5×7	Ship
10	200	320×256	3×3	Simulation

下载: 导出CSV

表 2 单帧数据库中6幅图像的详细信息

Table 2. Details of the six frames in the single-frame dataset

Signal image	Size	Target size	Target type
1	125×125	4×4	Plane
2	122×122	3×4	Plane
3	125×125	4×5	Plane
4	122×122	3×4	Plane
5	122×122	5×5	Plane
6	122×122	3×4	Unmanned vehicle

下载: 导出CSV

表 3 各图像序列下（均以其中一帧为例）不同算法的SCRG值

Table 3. The SCRG values of different algorithms for different image sequences

Seq	Target	DoG	ILCM	NLCM	RLCM	MPCM	WLDM	MDTDLMS	TCF	STLCF	Proposed
1	1	6.288	7.287	10.189	15.770	7.644	18.576	51.291	15.022	52.208	125.155
2	1	9.388	17.002	30.240	17.670	9.558	45.192	101.235	24.942	66.480	268.112
3	1	8.301	13.071	18.876	11.862	2.812	46.288	48.935	21.721	121.668	136.551
4	1	1.913	6.096	5.619	2.556	2.926	7.844	12.683	1.113	18.236	30.208
5	1	13.998	46.949	44.121	20.107	8.370	96.634	134.008	33.066	193.366	241.363
6	1	1.145	4.464	5.573	3.839	1.812	6.612	12.713	2.007	14.492	32.199
7	1	0.505	1.940	1.661	1.518	2.066	19.092	3.635	8.343	25.787	11.349
8	1	1.528	2.583	4.453	2.579	2.941	4.266	8.711	1.014	3.535	37.353
9	1	8.432	4.403	0.304	8.141	8.380	38.641	35.368	30.612	62.715	46.386
	2	6.086	14.072	7.662	7.071	6.465	34.149	44.157	3.869	3.957	44.898
	3	5.950	3.552	2.818	6.591	8.920	19.387	46.395	3.717	6.066	44.086
	4	5.331	3.422	0.065	6.591	6.974	18.155	37.250	4.195	7.411	43.760
10	1	1.123	2.023	2.638	7.106	2.680	0.746	0.675	7.374	12.176	378.726

下载: 导出CSV

表 4 各图像序列下（均以其中一帧为例）不同算法的BSF值

Table 4. The BSF values of different algorithms for different image sequences

Seq	DoG	ILCM	NLCM	RLCM	MPCM	WLDM	MDTDLMS	TCF	STLCF	Proposed
1	2.124	33.948	1.412	6.139	0.576	681.672	3.423E3	14.658	4.279	5.360E3
2	7.143	94.592	0.110	15.371	0.029	39.766	1.870E5	37.531	0.229	4.903E5
3	3.231	49.586	0.036	11.120	0.020	15.127	1.347E5	30.609	0.120	3.708E5
4	0.740	17.479	1.042	2.774	0.300	112.150	660.786	1.516	0.510	1.560E3
5	5.706	46.369	0.022	15.135	0.019	22.857	3.894E5	31.337	0.145	6.984E5
6	0.450	6.951	0.056	2.203	0.125	37.561	1.865E3	2.024	0.146	4.700E3
7	0.481	9.174	0.034	2.198	0.136	16.595	4.242E3	8.431	0.030	1.318E4
8	0.732	10.465	0.833	2.841	0.566	81.891	316.084	1.102	0.178	1.334E3
9	2.160	29.866	0.658	5.264	0.742	536.948	2.326E3	32.869	4.776	3.008E3
10	0.645	16.492	0.470	4.121	0.404	44.946	690.432	7.160	1.129	6.437E3

下载: 导出CSV

表 5 单帧图像库中不同算法的SCRG值

Table 5. The SCRG values of different algorithms for the single-frame image dataset

Frame	Target	DoG	ILCM	NLCM	RLCM	MPCM	WLDM	MDTDLMS	TCF	STLCF	Proposed
1	1	3.667	6.671	3.125	2.814	1.366	25.676	0.607	-	-	31.489
2	1	1.106	1.881	0.542	1.750	0.765	9.651	0.608	-	-	13.409
3	1	4.277	13.183	9.048	5.356	2.257	13.942	2.445	-	-	32.365
4	1	1.659	2.775	4.963	1.513	1.047	4.593	0.182	-	-	26.436
5	1	2.038	1.431	1.395	2.064	3.500	9.955	0.181	-	-	10.884
6	1	0.469	2.195	3.988	2.522	1.281	4.358	0.995	-	-	16.672

下载: 导出CSV

表 6 单帧图像库中不同算法的BSF值

Table 6. The BSF values of different algorithms for the single-frame image dataset

Frame	DoG	ILCM	NLCM	RLCM	MPCM	WLDM	MDTDLMS	TCF	STLCF	Proposed
1	1.630	5.904	0.065	2.742	0.235	58.382	172.100	-	-	2.699E3
2	0.332	2.677	0.034	1.360	0.260	14.742	93.599	-	-	1.268E3
3	1.127	11.319	0.142	2.856	0.148	32.569	357.666	-	-	3.024E3
4	0.558	9.072	0.318	1.528	0.292	23.423	90.231	-	-	1.036E3
5	0.635	4.898	0.082	1.513	0.358	29.906	60.236	-	-	536.425
6	0.199	3.619	0.025	1.308	0.118	4.992	268.685	-	-	2.743E3

下载: 导出CSV

表 7 各种图像序列下不同算法的运行时间

Table 7. Running times of different algorithms for different image sequences seconds/frame

Seq	DoG	ILCM	NLCM	RLCM	MPCM	WLDM	MDTDLMS	TCF	STLCF	Proposed
1	0.232	0.083	0.076	1.739	2.205	4.192	19.110	0.361	0.980	0.902
2	0.253	0.076	0.068	1.965	2.246	4.612	22.469	0.247	0.711	0.948
3	0.269	0.088	0.072	2.166	2.500	4.560	22.232	0.230	0.670	0.949
4	0.202	0.084	0.078	1.220	1.518	3.790	14.140	0.443	0.943	0.670
5	0.262	0.101	0.072	2.023	2.217	4.708	20.826	0.225	0.868	0.972
6	0.232	0.096	0.089	1.597	1.908	3.809	18.385	0.469	1.141	0.743
7	0.285	0.088	0.079	2.117	2.430	4.981	21.576	0.234	0.673	0.912
8	0.256	0.102	0.088	1.736	2.083	3.982	20.190	0.538	1.067	0.749
9	0.209	0.073	0.063	1.566	1.780	3.977	18.346	0.208	0.647	0.762
10	0.253	0.078	0.067	2.043	2.418	4.970	21.844	0.182	0.682	0.922

下载: 导出CSV

参考文献(18)

[1]	CUI Zheng, YANG Jingli, LI Junbao, et al. An infrared small target detection framework based on local contrast method[J]. Measurement, 2016, 91: 405-413. doi: 10.1016/j.measurement.2016.05.071
[2]	GAO Jinyan, LIN Zaiping, AN Wei. Infrared small target detection using a temporal variance and spatial patch contrast filter[J]. IEEE Access, 2019, 7: 32217-32226. doi: 10.1109/ACCESS.2019.2903808
[3]	BAI Xiangzhi, BI Yanguang. Derivative entropy-based contrast measure for infrared small-target detection [J]. IEEE Trans. Geosci. Remote Sens., 2018, 56(4): 2452-2466. doi: 10.1109/TGRS.2017.2781143
[4]	潘胜达, 张素, 赵明, 等. 基于双层局部对比度的红外弱小目标检测方法[J]. 光子学报, 2020, 49(1): 184-192. https://www.cnki.com.cn/Article/CJFDTOTAL-GZXB202001021.htm PAN Shengda, ZHANG Su, ZHAO Ming, et al. Infrared small target detection based on double-layer local contrast measure[J]. Acta Photonica Sinica, 2020, 49(1): 184-192. https://www.cnki.com.cn/Article/CJFDTOTAL-GZXB202001021.htm
[5]	WANG Xin, LV Guogang, XU Lizhong. Infrared dim target detection based on visual attention[J]. Infrared Physics & Technology, 2012, 55(6): 513-521. http://www.sciencedirect.com/science/article/pii/S1350449512000801
[6]	CHEN C L P, LI Hong, WEI Yantao, 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
[7]	HAN Jinhui, MA Yong, ZHOU Bo, 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
[8]	QIN Yao, LI Biao. Effective infrared small target detection utilizing a novel local contrast method[J]. IEEE Geoscience and Remote Sensing Letters, 2016, 13(12): 1890-1894. doi: 10.1109/LGRS.2016.2616416
[9]	王晓阳, 彭真明, 张萍, 等. 局部对比度结合区域显著性红外弱小目标检测[J]. 强激光与粒子束, 2015, 27(9): 32-38. https://www.cnki.com.cn/Article/CJFDTOTAL-QJGY201509007.htm WANG Xiaoyang, PENG Zhenming, ZHANG Ping, et al. Infrared small dim target detection based on local contrast combined with region saliency [J]. High Power Laser and Particle Beams, 2015, 27(9): 32-38. https://www.cnki.com.cn/Article/CJFDTOTAL-QJGY201509007.htm
[10]	WEI Yantao, YOU Xinge, LI Hong. Multiscale patch-based contrast measure for small infrared target detection [J]. Pattern Recognition, 2016, 58: 216-226. doi: 10.1016/j.patcog.2016.04.002
[11]	HAN Jinhui, LIANG Kun, ZHOU Bo, et al. Infrared small target detection utilizing the multi-scale relative local contrast measure[J]. IEEE Geoscience and Remote Sensing Letters, 2018, 15(4): 612-616. doi: 10.1109/LGRS.2018.2790909
[12]	DENG He, SUN Xianping, LIU Maili, et al. Entropy-based window selection for detecting dim and small infrared targets[J]. Pattern Recognition, 2017, 61: 66-77. doi: 10.1016/j.patcog.2016.07.036
[13]	KIM S, SUN S G, KIM K T. Highly efficient supersonic small infrared target detection using temporal contrast filter[J]. Electronics Letters, 2014, 50: 81-83. doi: 10.1049/el.2013.2109
[14]	DENG Lizhen, ZHANG Jieke, ZHU Hu. Infrared moving point target detection using a spatial-temporal filter[J]. Infrared Physics & Technology, 2018, 95: 122-127. http://www.sciencedirect.com/science/article/pii/S1350449518307916
[15]	MORADI S, MOALLEM P, SABAHI M F. Scale-space point function based framework to boost infrared target detection algorithms[J]. Infrared Physics & Technology, 2016, 77: 27-34. http://www.sciencedirect.com/science/article/pii/s1350449516300160
[16]	HAN Jinhui, LIU Sibang, QIN Gang, et al. A local contrast method combined with adaptive background estimation for infrared small target detection[J]. IEEE Geoscience and Remote Sensing Letters, 2019, 16(9): 1442-1446. doi: 10.1109/LGRS.2019.2898893
[17]	回丙伟, 宋志勇, 范红旗, 等. 地/空背景下红外图像弱小飞机目标检测跟踪数据集[J/OL]. 中国科学数据, 2020, 5(3): 286-297(DOI: 10.11922/csdata.2019.0074.zh). HUI Bingwei, SONG Zhiyong, FAN Hongqi, et al. A dataset for infrared detection and tracking of dim-small aircraft targets under ground/air background[J/OL]. China Scientific Data, 2020, 5(3): 286-297(DOI: 10.11922/csdata.2019.0074.zh).
[18]	OTCBVS WS Series Bench, Roland Miezianko, Terravic Research Infrared Database[DB/OL]. http://vcipl-okstate.org/pbvs/bench/index.html.