A camouflage identification method and system based on polarized light acquisition and band calculation

Abstract

The present invention relates to a target recognition method and system, and more specifically, to a camouflage recognition method and system based on polarized light acquisition and band calculation. This method addresses the technical issues that traditional spectral camouflage recognition requires confirmation of the spectral characteristics of the camouflage, multi-band spectral characteristics are difficult to identify, and camouflage and natural background often have different spectra but the same spectrum, making camouflage difficult to identify. The present invention's camouflage recognition method based on polarized light acquisition and band calculation can simultaneously acquire the target's polarized spectrum information, improving target recognition accuracy. The system includes a turntable, a polarized multispectral camera mounted on the turntable, an image processing module, and an image analysis and recognition module; the turntable and the polarized multispectral camera are respectively connected to an external control module; the polarized multispectral camera is connected to the input of the image processing module; and the output of the image processing module is connected to the image analysis and recognition module.

Description

Technical Field

[0001] The present invention relates to a target recognition method and system, and in particular to a camouflage recognition method and system based on polarized light acquisition and waveband calculation. Background Art

[0002] Target recognition plays a vital role in both daily life and national security. Improving target recognition accuracy is a key research area. The complex and ever-changing natural background, coupled with the diverse yet similar characteristics of targets, undoubtedly increases the difficulty of target recognition. Traditional multispectral recognition is significantly affected by the environment, has limited spectral bands, and captures little information. With the advancement of detection technology, multispectral recognition has gradually evolved to include spectral data, polarization data, and polarization-spectral data.

[0003] Polarization spectrum detection is a new ground detection method that has attracted widespread attention in recent years. Polarization state is an inherent property of electromagnetic waves. When objects on Earth or in the atmosphere reflect or scatter electromagnetic waves, they produce polarization characteristics determined by their state, and various target information is contained in these polarization characteristics. Polarization spectrum images contain multidimensional information such as spatial, spectral, and polarization information, combining the characteristics and advantages of spectral and polarization images. Polarization spectrum images contain more target information, which can increase the contrast between the target and the background, improve target detection capabilities, and are suitable for ground object detection in complex backgrounds.

[0004] The principle of camouflage is to use spectral camouflage to make the surface texture or color similar to the background. Therefore, camouflage and natural background often have the same spectrum, which makes it difficult to identify camouflage. Traditional spectral camouflage identification often requires confirming the spectral characteristics of the camouflage. Multi-band spectral characteristics are difficult to identify camouflage. Camouflage, as a man-made object, has different surface roughness and moisture content from the natural background, which makes the two significantly different in polarization characteristic parameters. Summary of the Invention

[0005] The present invention aims to address the technical issues that traditional spectral camouflage identification requires confirmation of the spectral characteristics of the camouflage, multi-band spectral characteristics are difficult to identify, and camouflage and natural background often have different spectra, making camouflage difficult to identify. Instead, a camouflage identification method and system based on polarization light acquisition and band calculation is provided, which can simultaneously obtain the polarization spectrum information of the target and improve the target recognition accuracy.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0007] A method for identifying camouflaged objects based on polarized light acquisition and band calculation is unique in that it includes the following steps:

[0008] Step 1: Obtain the original polarization spectrum image of the target scene;

[0009] Step 2: Separate and reconstruct the original polarization spectrum image to obtain polarization hyperspectral images corresponding to different polarization directions;

[0010] Step 3: Calculate the polarization parameters of the polarization hyperspectral image obtained in step 2;

[0011] Step 4: Calculate the NDVI index using the polarization parameters of the red band and the near-infrared band to obtain the NDVI index image;

[0012] Step 5: Binarize the NDVI index image to obtain a binary image;

[0013] Step 6: Perform corrosion and dilation operations on the binary image to obtain the target object and background after camouflage recognition.

[0014] Furthermore, step 1 is specifically as follows:

[0015] A polarization multispectral camera based on a linear gradient filter is fixed at the center of the rotation axis of a turntable. The number of displaced pixels s is determined, and the speed and rotation angle of the turntable are adjusted according to the number of displaced pixels s. The polarization multispectral camera is then used to obtain M original polarization spectral images of the target scene, where s is a multiple of 2 and M is an integer greater than or equal to 1.

[0016] Furthermore, step 2 is specifically as follows:

[0017] 2.1) Using a linear gradient filter, obtain wavelengths λ1, λ2, ..., λn from each original polarization spectrum image; extract the number s of displaced pixels corresponding to the response peak positions of wavelengths λ1, λ2, ..., λn, and concatenate the s columns of displaced pixels to obtain a single-band image containing the polarization direction, where n is an integer greater than or equal to 1;

[0018] The polarization directions include 0°, 45°, 90° and 135°;

[0019] 2.2) Decompose the single-band image obtained in step 2.1) according to the four polarization directions to obtain a polarization hyperspectral image.

[0020] Furthermore, step 3 is specifically as follows:

[0021] The polarization parameter S of the polarization hyperspectral image is calculated by the following formula:

[0022]

[0023] Where I represents the two polarization components I 0° , I 90°The sum of the light intensities at 0° and 90° is the total polarization intensity;

[0024] Q represents the two polarization components I 0° , I 90° The difference in light intensity at 0° and 90° respectively;

[0025] U represents the two polarization components I 45° , I 135° The sum of the light intensities at 45° and 135° respectively.

[0026] Furthermore, step 4 is specifically as follows:

[0027] The polarization parameters of the polarization hyperspectral images of the red band and the near-infrared band are arranged and combined, and the NDVI index images are calculated respectively. Then, the standard deviation of the NDVI index images is calculated, and the NDVI index image with the largest standard deviation is selected as the NDVI index image with the best effect.

[0028] The NDVI index is calculated using the following formula:

[0029]

[0030] Where: I RED Represents the red band, I NIR Represents the near-infrared band.

[0031] Furthermore, step 5 is specifically as follows:

[0032] According to the pixel values ​​on the NDVI index image calculated in step 4, draw the histogram corresponding to the NDVI index image, and set the minimum point on the curve corresponding to the histogram as the threshold T; binarize each pixel in the NDVI index image, set it to 0 when the NDVI index value is less than or equal to the threshold T, and set it to 255 when the NDVI index value is greater than the threshold T, to obtain a binary image.

[0033] Furthermore, step 6 is specifically as follows:

[0034] The pixels with a value of 0 on the binary image obtained in step 5 are used as graphic points, and the pixels with a value of 255 are used as background points. Erosion and dilation operations are performed to obtain an image after camouflage recognition.

[0035] The specific method of corrosion operation is:

[0036] Set the values ​​of all background points within the 3×3 neighborhood of the graphic point to 0;

[0037] The specific method of expansion operation is:

[0038] Set the values ​​of all graphic points within a 3×3 neighborhood of the background point to 255.

[0039] Furthermore, the method further includes step 7: displaying the image after the camouflage is identified through the image display module (5);

[0040] In step 2.1), the wavelengths λ1, λ2, ..., λn on each original polarization spectrum image are obtained by sequentially including the band information of 430 nm to 850 nm from left to right in each original polarization spectrum image, with a band interval of 5 nm, obtaining n = 85 bands;

[0041] In step 4, I RED The value of is 700nm, I NIR The value is 800nm.

[0042] At the same time, the present invention also provides a disguised object recognition system based on polarized light acquisition and band calculation for implementing the above-mentioned disguised object recognition method based on polarized light acquisition and band calculation, characterized by comprising a turntable, a polarized multispectral camera disposed on the turntable, an image processing module, and an image analysis and recognition module; the polarized multispectral camera is fixedly connected to the center of the rotating shaft of the turntable;

[0043] The turntable and the polarization multispectral camera are respectively connected to the external control module;

[0044] The polarization multispectral camera is connected to the input end of the image processing module; the output end of the image processing module is connected to the image analysis and recognition module;

[0045] An image processing module is used to separate and reconstruct the original polarization spectrum image acquired by the polarization multispectral camera;

[0046] The image analysis and recognition module is used to perform calculations and target recognition on the original polarization spectrum image.

[0047] Furthermore, it also includes an image display module connected to the image analysis and recognition module;

[0048] The image display module is used to display the image after the camouflage object is identified.

[0049] Compared with the prior art, the beneficial effects of the technical solution of the present invention are:

[0050] 1. The present invention's camouflaged object recognition method, based on polarized light acquisition and band operations, separates and reconstructs the original polarized spectral image. By calculating the polarization parameters of the red and near-infrared bands of the polarized hyperspectral image, the NDVI index is calculated and binarized. After binarization, erosion and dilation operations are performed. This method enables rapid recognition of camouflaged objects against natural backgrounds, improving the ability to identify camouflaged objects against natural backgrounds.

[0051] 2. The method of the present invention introduces spectral data such as polarization parameters, selects polarization hyperspectral images in four directions, and uses spectral data such as polarization parameters and NDVI index for calculation, thereby improving the distinguishability of target scenes and target backgrounds.

[0052] 3. The method of the present invention introduces the NDVI index for band calculation, uses the results of 700nm and 800nm ​​wavelengths for analysis, and uses the binary value method to segment the results. The calculation is simple and fast, and has high universal applicability.

[0053] 4. The present invention's camouflage identification system, based on polarized light acquisition and band calculation, utilizes the linear gradient filter and turntable of a polarized multispectral camera to overcome the shortcomings of traditional polarized multispectral cameras, such as limited image size and difficulty in stitching, and can capture multiple images in four polarization directions. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 The flowchart of the camouflage identification method based on polarized light acquisition and band calculation of the present invention is shown.

[0055] Figure 2 Schematic diagram of a single-band image obtained by separating and reconstructing an original polarization spectrum image in an embodiment of the present invention.

[0056] FIG3(a) is a schematic diagram of a single-band image including four polarization directions according to an embodiment of the present invention. Figure 1 .

[0057] FIG3( b ) is a schematic diagram of a single-band image including four polarization directions according to an embodiment of the present invention. Figure 2 .

[0058] FIG4( a ) is a schematic diagram of a single-band image at 0° in an embodiment of the present invention.

[0059] FIG4( b ) is a schematic diagram of a single-band image at 45° in an embodiment of the present invention.

[0060] FIG4( c ) is a schematic diagram of a single-band image at 90° in an embodiment of the present invention.

[0061] FIG4( d ) is a schematic diagram of a single-band image at 135° in an embodiment of the present invention.

[0062] Figure 5 Schematic diagram of the camouflage identification system based on polarized light acquisition and band calculation according to the present invention.

[0063] Figure 6 Schematic diagram of the process of obtaining single-band images in four polarization directions from an original polarization spectrum image in an embodiment of the present invention.

[0064] The accompanying drawings are denoted as follows:

[0065] 1-turntable, 2-polarization multispectral camera, 3-image processing module, 4-image analysis and recognition module, 5-image display module. DETAILED DESCRIPTION

[0066] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by ordinary technicians in this field based on the technical solution of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0067] like Figure 1 As shown, the present invention provides a method for identifying camouflage objects based on polarized light acquisition and band calculation, comprising the following steps:

[0068] Step 1: Adjust the rotation speed and rotation angle of the turntable 1 according to the number of displacement pixels to obtain the original polarization spectrum image of the target scene;

[0069] Determine the number of displaced pixels s, and adjust the rotation speed and rotation angle of the turntable 1 according to the number of displaced pixels s; then use the polarization multispectral camera 2 based on the linear gradient filter to obtain M original polarization spectral images of the target scene. In this embodiment, the value of s is preferably 8, and the value of M is preferably 1200; other technicians can also set s and M to other values ​​according to actual needs.

[0070] Step 2: Separate and reconstruct the original polarization spectrum image to obtain polarization hyperspectral images corresponding to different polarization directions;

[0071] 2.1) If Figure 2 As shown, a linear gradient filter is used to obtain the wavelength on each original polarization spectrum image. In this embodiment, 85 wavelengths are preferably obtained, specifically λ1, λ2, ..., λ 85 ; Extract wavelengths λ1, λ2, ..., λ according to the number of displacement pixels 8 85 The number of displacement pixels corresponding to the peak position of the response is 8 columns of pixels, and the 8 columns of pixels extracted at different wavelengths are used to obtain a single band image containing four polarization directions; the four polarization directions are 0°, 45°, 90° and 135°;

[0072] 2.2) If Figure 3(a) 、 3(b) As shown, the single-band image obtained in step 2.1) is decomposed according to the four polarization directions to obtain a polarization hyperspectral image.

[0073] Step 3: Calculate the polarization parameters of the polarization hyperspectral image obtained in step 2;

[0074] The 85 wavelengths λ1, λ2, ..., λ obtained in step 2.2) are 85 ; Polarization parameters of the polarization hyperspectral images under the corresponding four polarization directions are calculated. In this embodiment, the polarization parameter I in the Stokes vector is used as an example to calculate the polarization parameters of the polarization hyperspectral image, and the formula is:

[0075]

[0076] Where I represents the two polarization components I 0° , I 90° The sum of the light intensities at 0° and 90° is the total polarization intensity;

[0077] Q represents the two polarization components I 0° , I 90° The difference in light intensity at 0° and 90° respectively;

[0078] U represents the two polarization components I 45° , I 135° The sum of the light intensities at 45° and 135° respectively.

[0079] The calculated I parameter results are recorded as I1, I2, ..., I n .

[0080] Step 4: Calculate the NDVI index using the polarization parameters of the red band and the near-infrared band to obtain the NDVI index image;

[0081] Polarization parameters I1, I2, ..., I of polarization hyperspectral images in the red and near-infrared bands n Perform permutations and combinations, calculate the NDVI index images respectively, calculate the standard deviation of the NDVI index images, and select the NDVI index image with the largest standard deviation as the NDVI index image with the best effect; the corresponding best effect in this embodiment is the red band I RED 800nm ​​and near infrared band I NIR The polarization hyperspectral image at 700nm is used as the NDVI index image;

[0082] The NDVI index is calculated using the following formula:

[0083]

[0084] where R 800 、R 700 They represent the polarization hyperspectral images at 800nm ​​and 700nm respectively.

[0085] Step 5: Binarize the NDVI index image to obtain a binary image;

[0086] According to the pixel values ​​on the NDVI index image calculated in step 4, draw the histogram corresponding to the NDVI index image, and set the minimum point on the curve corresponding to the histogram as the threshold T; binarize each pixel in the NDVI index image, that is, when the NDVI index value is less than or equal to the threshold T, it is set to 0, and when the NDVI index value is greater than the threshold T, it is set to 255, to obtain a binary image.

[0087] Step 6: Perform corrosion and dilation operations on the binary image to obtain the target image after camouflage recognition.

[0088] The pixels with a value of 0 on the binary image obtained in step 5 are used as graphic points, and the pixels with a value of 255 are used as background points. Erosion and dilation operations are performed to obtain an image after camouflage recognition.

[0089] The specific method of corrosion operation is:

[0090] Set the values ​​of all background points within the 3×3 neighborhood of the graphic point to 0;

[0091] The specific method of expansion operation is:

[0092] Set the values ​​of all graphic points within a 3×3 neighborhood of the background point to 255.

[0093] Step 7: Display the image after the camouflage is identified through the image display module 5.

[0094] like Figure 5 As shown, at the same time, the present invention also provides a system for implementing the above-mentioned camouflage identification method based on polarized light acquisition and band operation, including a turntable 1, a polarization multispectral camera 2 set on the turntable 1, an image processing module 3 and an image analysis and recognition module 4;

[0095] The turntable 1 and the polarization multispectral camera 2 are respectively connected to the external control module; the polarization multispectral camera 2 is connected to the input end of the image processing module 3; the output end of the image processing module 3 is connected to the image analysis and recognition module 4; the image processing module 3 is used to separate and reconstruct the original polarization spectrum image obtained by the polarization multispectral camera 2; the image analysis and recognition module 4 is used to perform calculations and target recognition on the original polarization spectrum image.

[0096] First, the function of the turntable 1 is to control the rotation speed and angle of the polarization multispectral camera 2. The rotation speed is adjusted so that the frame rate of the image acquired by the polarization multispectral camera 2 matches the rotation speed. That is, the displacement between every two adjacent original polarization spectrum images is the same. The turntable 1 controls the total rotation angle to obtain the shooting range of the corresponding scene. The polarization multispectral camera 2 is fixed on the turntable 1.

[0097] The polarization multispectral camera 2 then captures raw polarization spectral images of the target scene at four polarization angles. The size of these images is determined by the angle of the turntable 1. The size of the original polarization-captured images is 2048 × 2448 × 1200, where 1200 represents the number of raw polarization spectral images captured. In this embodiment, 1200 images are captured. Each raw polarization spectral image contains information from the 430nm-850nm band from left to right, with a band spacing of 5nm, for a total of 85 bands. Each group of four pixels in the raw polarization spectral images contains components in the four polarization directions of 0°, 45°, 90°, and 135°. See FIG. Figure 4(a) to Figure 4(d) ;

[0098] Finally, the image processing module 3 is connected to the polarization multispectral camera 2 to obtain the original polarization spectrum image, separate it, and reconstruct it into a single spectral band image containing four polarization directions. The areas of the same band of 1200 images are extracted and spliced ​​to obtain a single spectral band image, each of which is 2048×1200×8 in size. Then, images of four polarization directions are separated from the single band image. Each single band image can be separated into four polarization hyperspectral images with different polarization directions. The polarization parameters of the single wavelength image are calculated using the polarization hyperspectral images with different polarization directions. Finally, the polarization parameters of each band are obtained, whose size is 1024×(1200×8 / 2).

[0099] In addition, an image display module 5 is provided, connected to the image analysis and recognition module 4; the image display module 5 is used to display the image after the disguised object is recognized. In this embodiment, the image display module 5 uses a liquid crystal display to display the final results. The image processing module 3 and the image analysis and recognition module 4 are integrated into a computer.

[0100] The working principle of the above embodiment is as follows:

[0101] like Figure 6 As shown, after the system of the present invention is powered on, the rotation speed and rotation angle of the turntable 1 are adjusted, and the original polarization spectrum image is obtained through the polarization multispectral camera 2. The polarization multispectral camera 2 is composed of a linear gradient filter and a polarization filter. The original polarization spectrum image obtained includes polarization directions at four angles of 0°, 45°, 90°, and 135°. The original polarization spectrum image obtained by shooting is input into the image processing module 3 to obtain a single band image including four polarization directions. The single band image is decomposed to obtain four polarization hyperspectral images. The polarization parameters of the polarization hyperspectral image are calculated, and polarization hyperspectral images corresponding to polarization parameters with wavelengths of 800nm ​​and 700nm are obtained therefrom. The polarization hyperspectral image is input into the image analysis and recognition module 4 to calculate the NDVI index image. The calculation formula is as follows:

[0102]

[0103] where R 800 、R 700 Represent the polarization hyperspectral images corresponding to 800nm ​​and 700nm respectively.

[0104] The acquired polarization hyperspectral image is binarized by selecting a threshold value, which can distinguish the target object and the background. Then, the binary image is eroded and expanded to remove small background recognition error points, improve the recognition accuracy, and reduce the false alarm rate. Finally, the recognition result is output to the image display module 5 for display.

Claims

1. A method for identifying camouflaged objects based on polarized light acquisition and band calculation, characterized in that: The following steps are involved: Step 1: Obtain the original polarization spectrum image of the target scene; Step 2: Separate and reconstruct the original polarization spectrum image to obtain polarization hyperspectral images corresponding to different polarization directions; Step 3: Calculate the polarization parameters of the polarization hyperspectral image obtained in step 2; Step 4: Calculate the NDVI index using the polarization parameters of the red band and the near-infrared band to obtain the NDVI index image; Step 5: Binarize the NDVI index image to obtain a binary image; Step 6: Perform corrosion and dilation operations on the binary image to obtain the target object and background after camouflage recognition.

2. The method for identifying camouflage objects based on polarized light acquisition and band calculation according to claim 1, characterized in that: Step 1 is as follows: A polarization multispectral camera (2) based on a linear gradient filter is fixed at the center position of the turntable (1), the number of displacement pixels s is determined, and the rotation speed and rotation angle of the turntable (1) are adjusted according to the number of displacement pixels s; and M original polarization spectrum images of the target scene are obtained using the polarization multispectral camera (2), where s is a multiple of 2 and M is an integer greater than or equal to 1.

3. The method for identifying camouflage objects based on polarized light acquisition and band calculation according to claim 2, characterized in that: Step 2 is as follows: 2.1) Using a linear gradient filter, obtain wavelengths λ1, λ2, ..., λn from each original polarization spectrum image; extract the number s of displaced pixels corresponding to the response peak positions of wavelengths λ1, λ2, ..., λn, and concatenate the s columns of displaced pixels to obtain a single-band image containing the polarization direction, where n is an integer greater than or equal to 1; The polarization directions include 0°, 45°, 90° and 135°; 2.2) Decompose the single-band image obtained in step 2.1) according to the four polarization directions to obtain a polarization hyperspectral image.

4. The method for identifying camouflage objects based on polarized light acquisition and band calculation according to claim 3, characterized in that: Step 3 is as follows: The polarization parameter S of the polarization hyperspectral image is calculated by the following formula: Where I represents the two polarization components I 0° , I 90° The sum of the light intensities at 0° and 90° is the total polarization intensity; Q represents the two polarization components I 0° , I 90° The difference in light intensity at 0° and 90° respectively; U represents the two polarization components I 45° , I 135° The sum of the light intensities at 45° and 135° respectively.

5. The method for identifying camouflage objects based on polarized light acquisition and band calculation according to claim 4, characterized in that: Step 4 is as follows: The polarization parameters of the polarization hyperspectral images of the red band and the near-infrared band are arranged and combined to calculate the NDVI index images respectively. The standard deviation of the NDVI index images is then calculated, and the NDVI index image with the largest standard deviation is selected as the NDVI index image with the best effect. The NDVI index is calculated by the following formula: Where: I RED Represents the red band, I NIR Represents the near-infrared band.

6. The method for identifying camouflage objects based on polarized light acquisition and band calculation according to claim 5, characterized in that: Step 5 is as follows: According to the pixel values ​​on the NDVI index image calculated in step 4, draw the histogram corresponding to the NDVI index image, and set the minimum point on the curve corresponding to the histogram as the threshold T; binarize each pixel in the NDVI index image, set it to 0 when the NDVI index value is less than or equal to the threshold T, and set it to 255 when the NDVI index value is greater than the threshold T, to obtain a binary image.

7. The method for identifying camouflage objects based on polarized light acquisition and band calculation according to claim 6, characterized in that: Step 6 is as follows: The pixels with a value of 0 on the binary image obtained in step 5 are used as graphic points, and the pixels with a value of 255 are used as background points. Erosion and dilation operations are performed to obtain an image after camouflage recognition. The specific method of the corrosion operation is: Set the values ​​of all background points within the 3×3 neighborhood of the graphic point to 0; The specific method of the expansion operation is: Set the values ​​of all graphic points within a 3×3 neighborhood of the background point to 255.

8. The method for identifying camouflaged objects based on polarized light acquisition and wavelength calculation according to claim 7, characterized in that: The method further includes step 7: displaying the image after the camouflage is identified through the image display module (5); In step 2.1), the wavelengths λ1, λ2, ..., λn on each original polarization spectrum image are obtained by sequentially including band information of 430 nm to 850 nm from left to right in each original polarization spectrum image, with a band interval of 5 nm, to obtain n = 85 bands; In step 4, the I RED The value of is 700nm, I NIR The value is 800nm.

9. A system for a disguised object recognition method based on polarized light acquisition and band calculation, for implementing the disguised object recognition method based on polarized light acquisition and band calculation as described in claims 1-8, characterized in that: The invention comprises a turntable (1), a polarization multispectral camera (2) arranged on the turntable (1), an image processing module (3), and an image analysis and recognition module (4); the polarization multispectral camera (2) is fixedly connected to the center position of the rotation axis of the turntable (1); The turntable (1) and the polarization multispectral camera (2) are respectively connected to an external control module; The polarization multispectral camera (2) is connected to the input end of the image processing module (3); the output end of the image processing module (3) is connected to the image analysis and recognition module (4); The image processing module (3) is used to separate and reconstruct the original polarization spectrum image acquired by the polarization multispectral camera (2); The image analysis and recognition module (4) is used for performing calculations and target recognition on the original polarization spectrum image.

10. The camouflage identification system based on polarized light acquisition and band calculation according to claim 9, characterized in that: It also includes an image display module (5) connected to the image analysis and recognition module (4); The image display module (5) is used to display the image after the camouflage object is identified.

Images