Intelligent suppression of infrared dim and small target detection method under complex space backgrounds

Fig. 1. A pipeline of infrared dim and small target detection method based on complex backgrounds intelligent suppression (The three-dimensional graph in the figure shows the grayscale value of each pixel in the image)

Download full size

View in Article

Fig. 2. (a) Infrared scene-optimized encoder-decoder background suppression network model; (b) Multi-level fusion mechanism; (c) Residual fusion module

Download full size

View in Article

Fig. 3. Residual unit structure diagram

Download full size

View in Article

Fig. 4. Infrared weak and small target simulation data in typical complex interference scenarios

Download full size

View in Article

Fig. 5. Background suppression and target detection results of 6 detection methods in complex interference scenario

Download full size

View in Article

Fig. 6. Target detection results of the proposed method in various complex interference scenarios. (a) Original images; (b) Ground-truth images

Download full size

View in Article

Table 1. Infrared scene optimization encoding and decoding background suppression network model parameters

View table

View in Article

Table 1. Infrared scene optimization encoding and decoding background suppression network model parameters

Layers	Filter size	Filter number	Stride
Input	-	-	-
Conv1	7×7	32	1
Conv2	3×3	64	2
Conv3	3×3	128	2
RFM	$ \left[\begin{array}{l}3 \times 3 \\3 \times 3\end{array}\right] \times 8 $	$ \left[\begin{array}{l}64 \\64\end{array}\right] \times 8 $	$ \left[\begin{array}{l}1 \\1\end{array}\right] \times 8 $
Deconv1	3×3	64	2
Mixing layer 1	Conv2+Deconv1
Deconv2	3×3	32	2
Mixing layer 2	Conv1+Deconv2
Conv22	3×3	1	1
Residual offset	Input-Conv22
Output	-	-	-

Table 2. Three groups of test data parameters
View table
View in Article
Table 2. Three groups of test data parameters
Sequences Number of frames Image resolution Target size
1 600 256×256 2×2
2 675 256×256 2×2
3 800 256×256 2×2
4 620 256×256 2×2
5 700 256×256 2×2
6 600 256×256 2×2

Table 3. Background standard deviation \begin{document}$ {STD}_{{\mathrm{B}}} $\end{document} on six methods in six scenarios

View table

View in Article

Table 3. Background standard deviation \begin{document}$ {STD}_{{\mathrm{B}}} $\end{document} on six methods in six scenarios

Method	Sequence 1	Sequence 2	Sequence 3	Sequence 4	Sequence 5	Sequence 6
Original images	20.1463	20.0492	20.1374	20.0906	16.8078	105.1225
Sobel gradient	165.8147	169.3257	176.5302	186.1344	140.0292	182.2332
Median filtering	7.3690	7.3644	7.3691	7.3666	6.8255	40.2364
Bilateral filtering	7.5342	7.5335	7.5453	7.5524	6.7268	43.1227
Morphological	6.3837	6.3811	6.3840	6.3819	6.2816	48.2588
RLCM	5.6526	5.3424	5.5635	5.4432	5.2886	28.5535
Our method	1.8450	1.8500	1.8554	1.8639	2.9927	22.4826

Table 4. Target signal-to-noise ratio \begin{document}$ SNR $\end{document} on six methods in six scenarios

View table

View in Article

Table 4. Target signal-to-noise ratio \begin{document}$ SNR $\end{document} on six methods in six scenarios

Method	Sequence 1	Sequence 2	Sequence 3	Sequence 4	Sequence 5	Sequence 6
Original images	2.2686	4.8347	7.3604	9.9219	6.6991	1.8745
Sobel gradient	3.9824	9.1420	13.8655	18.0029	10.1626	2.3422
Median filtering	5.7071	12.4372	19.1936	26.0031	15.5626	6.6526
Bilateral filtering	6.2456	12.4962	18.7207	24.9382	15.8818	6.8235
Morphological	7.3883	15.2069	23.0435	30.8808	16.2214	6.5622
RLCM	8.0865	16.9081	29.6414	28.6213	20.5213	8.2312
Our method	25.1897	54.2541	82.7290	110.0257	29.9623	10.7003

Table 5. Detection rates \begin{document}$ {P}_{{\mathrm{d}}} $\end{document} on six methods in six scenarios

View table

View in Article

Table 5. Detection rates \begin{document}$ {P}_{{\mathrm{d}}} $\end{document} on six methods in six scenarios

Method	Sequence 1	Sequence 2	Sequence 3	Sequence 4	Sequence 5	Sequence 6
Sobel gradient	0.2611	0.4233	0.5237	0.7011	0.5236	0.4023
Median filtering	0.9041	0.8955	0.8758	0.9195	0.8213	0.7852
Bilateral filtering	0.9650	0.9602	0.9844	0.9652	0.8528	0.8021
Morphological	0.9652	0.9662	0.9480	0.9532	0.8467	0.8215
RLCM	0.9688	0.9702	0.9624	0.9708	0.9017	0.8826
Our method	0.9986	0.9987	0.9988	0.9989	0.9486	0.9125

Table 6. False alarm rates \begin{document}$ {F}_{{\mathrm{a}}} $\end{document} on six methods in six scenarios

View table

View in Article

Table 6. False alarm rates \begin{document}$ {F}_{{\mathrm{a}}} $\end{document} on six methods in six scenarios

Method	Sequence 1	Sequence 2	Sequence 3	Sequence 4	Sequence 5	Sequence 6
Sobel gradient	6.1824	4.5195	0.9115	0.2612	4.8214	5.2369
Median filtering	4.8965	3.7582	0.6646	0.1387	3.2592	3.6245
Bilateral filtering	3.5145	2.5194	0.6224	0.1310	2.8688	3.6387
Morphological	4.2571	3.0348	0.4524	0.0560	2.1822	2.8900
RLCM	3.2375	2.8522	0.3218	0.0628	2.1536	1.8249
Our method	0.1824	0.0882	0.0207	0.0013	0.0856	0.9570

Tools

Get Citation

Copy Citation Text

Shiran GE, Ruikai LIU, Na LI, Huijie ZHAO. Intelligent suppression of infrared dim and small target detection method under complex space backgrounds[J]. Infrared and Laser Engineering, 2024, 53(11): 20240290

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: 图像处理

Received: Jul. 4, 2024

Accepted: --

Published Online: Dec. 13, 2024

The Author Email: LI Na (lina_17@buaa.edu.cn)

DOI:10.3788/IRLA20240290

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Infrared scene optimization encoding and decoding background suppression network model parameters

Table 1. Infrared scene optimization encoding and decoding background suppression network model parameters

Table 2. Three groups of test data parameters

Table 2. Three groups of test data parameters

Table 3. Background standard deviation \begin{document}$ {STD}_{{\mathrm{B}}} $\end{document} on six methods in six scenarios

Table 3. Background standard deviation \begin{document}$ {STD}_{{\mathrm{B}}} $\end{document} on six methods in six scenarios

Table 4. Target signal-to-noise ratio \begin{document}$ SNR $\end{document} on six methods in six scenarios

Table 4. Target signal-to-noise ratio \begin{document}$ SNR $\end{document} on six methods in six scenarios

Table 5. Detection rates \begin{document}$ {P}_{{\mathrm{d}}} $\end{document} on six methods in six scenarios

Table 5. Detection rates \begin{document}$ {P}_{{\mathrm{d}}} $\end{document} on six methods in six scenarios

Table 6. False alarm rates \begin{document}$ {F}_{{\mathrm{a}}} $\end{document} on six methods in six scenarios

Table 6. False alarm rates \begin{document}$ {F}_{{\mathrm{a}}} $\end{document} on six methods in six scenarios