A multispectral camouflage recognition imaging system and a camouflage recognition method

Abstract

The present application relates to camouflage recognition imaging system and camouflage recognition method, specifically relates to a kind of multispectral camouflage recognition imaging system and camouflage recognition method, solve the technical problems that existing camouflage recognition imaging system range of matter is small, application scene is limited, or camouflage recognition method cannot real-time imaging, real-time identification, or spectral database construction efficiency is low.The multispectral camouflage recognition system of the present application includes image acquisition unit, data processing unit and display storage unit, image acquisition unit uses snapshot type spectral imaging chip, can real-time exposure, real-time data processing;Data processing unit can obtain continuous spectral information by means of spectral fitting degree identification technology, according to the characteristic spectrum of different kinds of identification target selects feature database, changes identification target, greatly increases the identifiable spectral range.In the multispectral camouflage recognition method of the present application, spectral correction is included, can remove the interference of environmental light factor, work under different daylight color temperature.

Description

TECHNICAL FIELD

[0001] The present application relates to a camouflage recognition imaging system and a camouflage recognition method, in particular to a multispectral camouflage recognition imaging system and a camouflage recognition method. BACKGROUND

[0002] The conventional camouflage recognition imaging system mainly relies on the specific spectral reflectivity, color and other characteristics of the camouflage recognition target. Such a camouflage recognition imaging system has a single purpose, and only one imaging system can recognize the camouflage of a specific number of substances based on fixed characteristics. To recognize multiple targets, multiple imaging systems with different characteristics need to be built-in, which is costly.

[0003] The current mature camouflage recognition method based on multispectral imaging requires multiple exposures to construct a spectral data cube. Therefore, when using a drone to carry a multispectral imaging system for imaging recognition, continuous shooting is usually required, and the scene must remain relatively static during shooting. This makes it difficult for the camouflage recognition method to adapt to the needs of complex and variable anti-camouflage scenes, severely limiting its application scenarios. In addition, the application scenarios of the camouflage recognition method are mostly outdoor daylight environments. The daylight color temperature is different at different times in different weather, and the reflection spectrum of the object is different. Therefore, it is necessary to construct a spectral database under different color temperatures, which requires a large amount of samples and has low construction efficiency.

[0004] Therefore, there is an urgent need for a multispectral imaging system that can real-time imaging and real-time recognition, and a camouflage recognition method adapted thereto to solve the above problems. SUMMARY

[0005] The purpose of the present application is to solve the technical problems of the existing camouflage recognition imaging system, which has a small range of recognized substances, limited application scenarios, or the camouflage recognition method cannot real-time imaging and real-time recognition, or the spectral database construction efficiency is low, and to provide a multispectral camouflage recognition imaging system and a camouflage recognition method.

[0006] To achieve the above purpose, the technical solution adopted by the present application is:

[0007] A multispectral camouflage recognition imaging system, characterized in that it comprises a calibration whiteboard, an image acquisition unit and a data processing unit connected by electricity, and a display storage unit connected with the data processing unit by electricity;

[0008] The calibration whiteboard is located in the shooting area of the camouflage recognition target during calibration;

[0009] The image acquisition unit comprises a band-pass filter, an imaging lens group and a snapshot spectral imaging chip arranged in sequence along the light path of the camouflage recognition target, and a sensor circuit group connected with the snapshot spectral imaging chip;

[0010] The data processing unit comprises a spectral reconstruction module, a true color restoration module, a color recognition module, a spectral classification recognition module and an identification result output module connected in sequence, and a spectral correction module;

[0011] The input end of the spectral correction module is connected with the output end of the spectral reconstruction module, and the output end is connected with the input end of the spectral classification recognition module;

[0012] The snapshot spectral imaging chip is electrically connected with the spectral reconstruction module through a sensor circuit group, and the identification result output module is electrically connected with a display storage unit;

[0013] The true color restoration module is configured with a CIE-RGB1931 chromaticity curve;

[0014] The color recognition module is configured with an RGB color database, which comprises standard RGB channel data of multiple camouflage identification targets;

[0015] The spectral classification recognition module is configured with a spectral reflectivity database, which comprises camouflage object reflectivity spectra of multiple approximate target camouflage objects and camouflage material reflectivity spectra of multiple approximate target camouflage materials.

[0016] Further, the data processing unit is an external computer or an internal FPGA.

[0017] Further, the display storage unit comprises a digital image display device and a data storage device;

[0018] The digital image display device and the data storage device are respectively electrically connected with the identification result output module.

[0019] A multispectral camouflage identification method based on the multispectral camouflage identification imaging system, which is characterized by comprising the following steps:

[0020] Step 1, obtaining a spectral reflectivity database comprising standard spectra and configuring it in the spectral classification recognition module; the standard spectra comprise camouflage object reflectivity spectra of multiple approximate target camouflage objects and camouflage material reflectivity spectra of multiple approximate target camouflage materials;

[0021] Step 2, obtaining an RGB color database comprising standard RGB channel data and configuring it in the color recognition module; the RGB color database comprises standard RGB channel data of multiple camouflage identification targets;

[0022] Step 3, placing a calibration whiteboard in the shooting area of the camouflage identification target, and the image acquisition unit obtains the spectral image of the calibration whiteboard, and the spectral reconstruction module obtains the environmental spectrum of the calibration whiteboard;

[0023] Step 4: Remove the calibration whiteboard. The image acquisition unit acquires a grayscale image of the camouflage recognition target shooting area. The spectral reconstruction module converts the grayscale image into spectral data cubes.

[0024] Step 5: The true color restoration module converts the spectral data cubically into an RGB image;

[0025] Step 6: The color recognition module selects the corresponding standard RGB channel data from the RGB color database configured in Step 2 based on the camouflage recognition target, and performs preliminary color filtering on the RGB image obtained in Step 5 based on the standard RGB channel data to obtain the camouflage recognition pixel area with the color of the camouflage recognition target.

[0026] Step 7: Select the spectral data cube corresponding to the camouflage recognition pixel area from the spectral data cube obtained in Step 4. The spectral correction module uses the environmental spectrum obtained in Step 3 to correct the spectral data cube corresponding to the camouflage recognition pixel area, and obtains the corrected spectrum of each pixel in the camouflage recognition pixel area.

[0027] Step 8: The spectral classification and recognition module selects the corresponding reflectance spectrum of the camouflage object and the reflectance spectrum of the camouflage material from the spectral reflectance database configured in Step 1, based on the camouflage identification target. It then compares the corrected spectrum of each pixel within the camouflage identification pixel area with the reflectance spectrum of the camouflage object and the reflectance spectrum of the camouflage material. If a pixel contains the spectral features of the camouflage identification target, that pixel is selected; otherwise, the comparison continues to the next pixel. The area composed of pixels containing the spectral features of the camouflage identification target within the camouflage identification pixel area is recorded as the camouflage pixel area.

[0028] Step 9: Mark the spoofed pixel regions in the RGB image obtained in step 5 to obtain a color image with spoofed markings. Then, send the color image with spoofed markings to the display storage unit through the recognition result output module. The display storage unit displays and stores the color image with spoofed markings to complete the spoofing recognition.

[0029] Furthermore, step 8 specifically includes:

[0030] Step 8.1: The spectral classification and recognition module selects the corresponding camouflage object reflectance spectrum and camouflage material reflectance spectrum from the spectral reflectance database configured in Step 1, based on the camouflage identification target.

[0031] Step 8.2: Select the characteristic spectral band of the corrected spectrum of each pixel in the camouflage recognition pixel area, and normalize the spectrum of the corresponding pixel using the characteristic spectral band as the standard to obtain the normalized spectrum of each pixel in the camouflage recognition pixel area.

[0032] Select the characteristic spectral bands of the reflectance spectrum of the camouflage object and the characteristic spectral bands of the reflectance spectrum of the camouflage material, and normalize the reflectance spectra of the camouflage object and the camouflage material respectively using the characteristic spectral bands as the standard to obtain the normalized reflectance spectra of the camouflage object and the normalized reflectance spectra of the camouflage material.

[0033] Step 8.3: Calculate the Euclidean distance between the normalized spectrum of each pixel in the characteristic spectral band and the normalized reflectance spectrum of the camouflage object, and the Euclidean distance between the normalized spectrum of each pixel and the reflectance spectrum of the camouflage material, to obtain two sets of Euclidean distances for each pixel.

[0034] Step 8.4: Set the spectral recognition threshold. Compare the two sets of Euclidean distances for each pixel obtained in Step 8.3 with the spectral recognition threshold. If both sets of Euclidean distances for a pixel are greater than the spectral recognition threshold, then classify the pixel as "other".

[0035] If at least one set of Euclidean distances of a pixel is less than or equal to the spectral recognition threshold, and the Euclidean distance between the normalized spectrum of the pixel and the reflectance spectrum of the camouflage object is less than the Euclidean distance between the normalized spectrum and the reflectance spectrum of the camouflage material, then the pixel is classified as a camouflage object.

[0036] If at least one set of Euclidean distances of a pixel is less than or equal to the spectral recognition threshold, and the Euclidean distance between the normalized spectrum of the pixel and the reflectance spectrum of the camouflage object is greater than the Euclidean distance between the normalized spectrum and the reflectance spectrum of the camouflage material, then the pixel is classified as a camouflage material.

[0037] Step 8.5: Mark the pixels that are classified as camouflage materials. The area composed of such pixels is the camouflage pixel area.

[0038] Furthermore, step 6 specifically includes:

[0039] Step 6.1: The color recognition module selects the corresponding standard RGB channel data from the RGB color database configured in step 2 according to the disguised recognition target. The standard RGB channel data is normalized using the gray value of channel G in the standard RGB channel data as the standard to obtain normalized standard RGB channel data. The normalized standard RGB channel data includes standard R channel gray value data, standard G channel gray value data and standard B channel gray value data.

[0040] Step 6.2: Extract the RGB channel data of each pixel in the RGB image obtained in Step 5. According to the normalization method in Step 6.1, normalize the RGB channel data with the gray value of channel G in the RGB channel data of each pixel as the standard to obtain the normalized RGB channel data of each pixel. The normalized RGB channel data includes R channel gray value data, G channel gray value data and B channel gray value data.

[0041] Step 6.3: Calculate the Euclidean distance between the normalized RGB channel data and the normalized standard RGB channel data for each pixel using the following formula:

[0042]

[0043] Where L represents the Euclidean distance between the normalized RGB channel data and the normalized standard RGB channel data, the smaller the Euclidean distance, the smaller the difference between the two sets of data;

[0044] r0, g0, and b0 represent standard R-channel grayscale data, standard G-channel grayscale data, and standard B-channel grayscale data, respectively.

[0045] r, g, and b represent the grayscale data of the R channel, G channel, and B channel of a pixel in an RGB image, respectively.

[0046] Step 6.4: Set the color recognition threshold. Compare the Euclidean distance between the normalized RGB channel data and the normalized standard RGB channel data of each pixel with the color recognition threshold. Pixels whose normalized RGB channel data have an Euclidean distance less than the color recognition threshold are recorded as target pixels. The area composed of target pixels is the camouflage recognition pixel area with camouflage recognition target color.

[0047] Furthermore, in step 6.1, the specific method for normalization is as follows:

[0048] A. Divide the grayscale value of channel G in the standard RGB channel data by 1 to obtain the coefficient a;

[0049] B. Divide the gray values ​​r1 of channel R, g1 of channel G, and b1 of channel B in the standard RGB channel data by the coefficient a to obtain the normalized channel gray values ​​r, g, and b respectively.

[0050] C. Integrate r, g, and b into normalized standard RGB channel data.

[0051] Furthermore, step 5 specifically includes:

[0052] Step 5.1: Multiply the spectral dimension matrix of each pixel in the spectral data cube obtained in Step 4 by the CIE-RGB1931 chromaticity coordinate curve to obtain the RGB channel grayscale value of each pixel:

[0053] RGB(x,y) = CIE*S(x,y)

[0054] Where S(x,y) represents the spectral dimension matrix of the pixel at spatial coordinate (x,y), RGB(x,y) represents the RGB channel grayscale value of the pixel at spatial coordinate (x,y), and CIE represents the CIE-RGB1931 chromaticity coordinate curve.

[0055] Step 5.2: Arrange the gray values ​​of the RGB channels of each pixel in the order of the cubic space coordinates of the spectral data to obtain the RGB image.

[0056] Further, step 1 specifically includes:

[0057] Step 1.1: Obtain the linear differences in characteristic spectral bands of approximate target camouflage objects and approximate target camouflage materials, and prepare a certain number of approximate target camouflage objects and approximate target camouflage materials as samples for constructing the spectral reflectance database;

[0058] Step 1.2: Place the calibration whiteboard in sunlight, the image acquisition unit acquires the spectral image of the calibration whiteboard, and the spectral reconstruction module obtains the sunlight environment spectrum;

[0059] Step 1.3: The image acquisition unit acquires spectral images of the approximate target camouflage object and the approximate target camouflage material, respectively, and uses the spectral reconstruction module to obtain the spectrum of the camouflage object and the spectrum of the camouflage material;

[0060] Step 1.4: The spectral correction module corrects the camouflage object spectrum and camouflage material spectrum obtained in Step 1.3 based on the solar environment spectrum obtained in Step 1.2, to obtain the reflectance spectrum of the camouflage object and the reflectance spectrum of the camouflage material. The characteristic spectral segments of the reflectance spectrum of the camouflage object and the reflectance spectrum of the camouflage material are selected and recorded.

[0061] Step 1.5: Replace the approximate target camouflage object and approximate target camouflage material, and repeat steps 1.3-1.4 to obtain multiple sets of characteristic spectral bands of the reflectance spectra of the camouflage object and the reflectance spectra of the camouflage material for all the approximate target camouflage objects and approximate target camouflage materials prepared in step 1.1;

[0062] Step 1.6: Determine whether the characteristic spectral segments of the multiple sets of reflectance spectra of camouflage objects and reflectance spectra of camouflage materials obtained in Step 1.5 exhibit the linear difference pattern of the characteristic spectral segments obtained in Step 1.1, thereby verifying their effectiveness and obtaining effective characteristic spectral segments of reflectance spectra of camouflage objects and reflectance spectra of camouflage materials;

[0063] Step 1.7: Store the reflectance spectra of the camouflage object and the reflectance spectra of the camouflage material corresponding to the characteristic spectral segments of the effective camouflage object reflectance spectrum and the camouflage material reflectance spectrum as standard spectra in the spectral reflectance database, and configure the spectral reflectance database into the spectral classification and recognition module.

[0064] Furthermore, step 2 specifically includes:

[0065] Step 2.1: Place the camouflage target under sunlight and use a colorless RGB color camera to record the RGB three-channel grayscale information of the camouflage target;

[0066] Step 2.2: Change the disguised identification target and repeat step 2.1 to obtain the RGB three-channel grayscale information of various disguised identification targets. The RGB three-channel grayscale information of various disguised identification targets is composed of standard RGB channel data.

[0067] Step 2.3: Store the standard RGB channel data in the RGB color database and configure the RGB color database into the color recognition module.

[0068] Compared with the prior art, the present invention has the following beneficial technical effects:

[0069] 1. The multispectral camouflage recognition imaging system provided by the present invention uses a snapshot-type spectral imaging chip, which gives the imaging system the advantages of real-time exposure and real-time data processing, and enables continuous recognition in motion scenes;

[0070] 2. In the multispectral camouflage recognition imaging system provided by the present invention, the spectral classification and recognition module obtains continuous spectral information by means of spectral fitting degree recognition technology;

[0071] 3. In the multispectral camouflage recognition imaging system provided by the present invention, the color recognition module is equipped with an RGB color database, and the spectral classification recognition module is equipped with a spectral reflectance database. The feature database can be switched according to the spectral characteristics of different types of recognition targets, thereby changing the recognition target. No hardware adjustments are required. With a single detector, functions such as color image imaging, camouflage recognition, and area marking can be realized simultaneously. The camouflage area is clearly displayed while also having a low cost.

[0072] 4. The multispectral camouflage identification method provided by this invention includes a spectral correction module, which can remove the interference of ambient light factors and work under different daylight color temperatures. This can reduce the sample size required to build the spectral reflectance database and improve the efficiency of building the spectral reflectance database. Attached Figure Description

[0073] Figure 1This is a schematic diagram of an embodiment of the multispectral camouflage recognition imaging system of the present invention;

[0074] Figure 2 This is a schematic diagram of the structure of an embodiment of the multispectral camouflage recognition imaging system of the present invention;

[0075] The annotations in the attached figures are explained as follows:

[0076] 101-Image acquisition unit, 111-Bandpass filter, 121-Imaging lens group, 131-Snapshot type spectral imaging chip, 132-Sensor circuit group, 141-Calibration whiteboard, 142-Calibration whiteboard storage fixture;

[0077] 201-Data Processing Unit, 211-Spectral Reconstruction Module, 221-True Color Restoration Module, 231-Spectral Correction Module, 241-Color Recognition Module, 251-RGB Color Database, 261-Spectral Classification and Recognition Module, 271-Spectral Reflectance Database, 281-Recognition Result Output Module;

[0078] 301 - Display storage unit; 311 - Digital image display device; 321 - Data storage device. Detailed Implementation

[0079] The multispectral camouflage identification imaging system and camouflage identification method proposed in this invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Those skilled in the art should understand that these embodiments are merely used to explain the technical principles of this invention and are not intended to limit the scope of protection of this invention.

[0080] A multispectral camouflage identification imaging system, such as Figure 1 , Figure 2 As shown, it includes a calibration whiteboard 141, an image acquisition unit 101 and a data processing unit 201 electrically connected to each other, and a display storage unit 301 electrically connected to the data processing unit 201.

[0081] The calibration whiteboard 141 is located in the shooting area of ​​the camouflaged target identification device during calibration. In this embodiment, a calibration whiteboard storage fixture 142 can also be provided for storing the calibration whiteboard 141. Except during calibration, the calibration whiteboard 141 is stored in the calibration whiteboard storage fixture 142. The calibration whiteboard 141 can calibrate the light source spectrum at any time to obtain the reflectance spectrum of the material.

[0082] The image acquisition unit 101 includes a bandpass filter 111, an imaging lens group 121, and a snapshot-type spectral imaging chip 131 arranged sequentially along the optical path of the camouflage-identifying target, as well as a sensor circuit group 132 connected to the snapshot-type spectral imaging chip 131. The snapshot-type spectral imaging chip 131 can acquire mosaic images and has real-time exposure and real-time data processing functions.

[0083] The data processing unit 201 includes a spectral reconstruction module 211, a true color restoration module 221, a color recognition module 241, a spectral classification and recognition module 261, and a recognition result output module 281, all connected in sequence, as well as a spectral correction module 231. The input of the spectral correction module 231 is connected to the output of the spectral reconstruction module 211, and its output is connected to the input of the spectral classification and recognition module 261. The true color restoration module 221 is equipped with a CIE-RGB1931 chromaticity curve, which can convert spectral data into RGB three-channel grayscale data. The color recognition module 241 is equipped with an RGB color database 251, which includes standard RGB channel data for multiple camouflage targets. The spectral classification and recognition module 261 is equipped with a spectral reflectance database 271, which includes the reflectance spectra of camouflage objects and camouflage materials for multiple approximate target camouflage objects.

[0084] The data processing unit 201 is an external computer or an internal FPGA. Utilizing spectral fitting recognition technology, it can acquire continuous spectral information. By switching the feature database according to the spectral characteristics of different target types, the target can be changed, greatly increasing the identifiable spectral range. The spectral reconstruction module 211 incorporates an algorithm compatible with the snapshot-type spectral imaging chip 131, which converts the mosaic image into a spectral data cube. To eliminate spectral variations caused by changes in the light source, a spectral correction module 231 is added to the data processing unit 201. Spectral correction requires integration with light source spectral calibration data. The RGB color database 251 is pre-constructed based on the target color, and the spectral reflectance database 271 is pre-constructed based on the target's spectral characteristics.

[0085] The display storage unit 301 includes a digital image display device 311 and a data storage device 321. The snapshot-type spectral imaging chip 131 is electrically connected to the spectral reconstruction module 211 through the sensor circuit group 132. The recognition result output module 281 is electrically connected to the digital image display device 311 and the data storage device 321, respectively. The digital image display device 311 is an integrated display, and the data storage device 321 is a built-in memory. In other embodiments, the display storage unit 301 can also be an external computer.

[0086] The data processing unit 201 can be integrated into an electronic device to complete data reading, processing, and storage. When operating, the data processing unit 201 needs to read the pre-selected characteristic spectrum and RGB database based on the target being identified, as well as the whiteboard calibration image of the working environment. The image acquisition unit 201 records the grayscale information of the target area to be measured, and the data processing unit 201 performs data processing and analysis processes such as spectral reconstruction to complete the camouflage identification of the target area. The identification result is stored and displayed by the display storage unit 301. The data processing unit 201 and the display storage unit 301, with the help of the true color restoration module 221, can generate a color image with the identification result, making operation simple.

[0087] A multispectral camouflage identification method, based on the aforementioned multispectral camouflage identification imaging system, is characterized by comprising the following steps:

[0088] Step 1: Obtain a spectral reflectance database 271 including standard spectra and configure it into the spectral classification and recognition module 261. The standard spectra include the reflectance spectra of multiple approximate target camouflage objects and the reflectance spectra of multiple approximate target camouflage materials. The specific method is as follows:

[0089] Step 1.1: Obtain the linear differences in characteristic spectral bands of approximate target camouflage objects and approximate target camouflage materials, and prepare a certain number of approximate target camouflage objects and approximate target camouflage materials as samples for constructing the spectral reflectance database 271;

[0090] Step 1.2: Place the calibration whiteboard 141 under sunlight, and the image acquisition unit 101 acquires the spectral image of the calibration whiteboard 141, and uses the spectral reconstruction module 211 to obtain the sunlight environment spectrum.

[0091] Step 1.3: The image acquisition unit 101 acquires the spectral images of the approximate target camouflage object and the approximate target camouflage material respectively, and uses the spectral reconstruction module 211 to obtain the spectrum of the camouflage object and the spectrum of the camouflage material;

[0092] Step 1.4: The spectral correction module 231 corrects the camouflage object spectrum and camouflage material spectrum obtained in Step 1.3 based on the solar environment spectrum obtained in Step 1.2, and obtains the reflectance spectrum of the camouflage object and the reflectance spectrum of the camouflage material. The characteristic spectral segments of the reflectance spectrum of the camouflage object and the reflectance spectrum of the camouflage material are selected and recorded.

[0093] Step 1.5: Replace the approximate target camouflage object and approximate target camouflage material, and repeat steps 1.3-1.4 to obtain multiple sets of characteristic spectral bands of the reflectance spectra of the camouflage object and the reflectance spectra of the camouflage material for all the approximate target camouflage objects and approximate target camouflage materials prepared in step 1.1;

[0094] Step 1.6: Determine whether the characteristic spectral segments of the multiple sets of reflectance spectra of camouflage objects and reflectance spectra of camouflage materials obtained in Step 1.5 exhibit the linear difference pattern of the characteristic spectral segments obtained in Step 1.1, thereby verifying their effectiveness and obtaining effective characteristic spectral segments of reflectance spectra of camouflage objects and reflectance spectra of camouflage materials;

[0095] Step 1.7: Store the reflectance spectra of the camouflage object and the reflectance spectra of the camouflage material corresponding to the characteristic spectral segments of the effective camouflage object reflectance spectrum and the camouflage material reflectance spectrum as standard spectra in the spectral reflectance database 271, and configure the spectral reflectance database 271 into the spectral classification and recognition module 261.

[0096] Step 2: Obtain the RGB color database 251, which includes standard RGB channel data, and configure it into the color recognition module 241; wherein, the RGB color database 251 includes standard RGB channel data of multiple disguised recognition targets. The specific method is as follows:

[0097] Step 2.1: Place the camouflage target under sunlight and use a colorless RGB color camera to record the RGB three-channel grayscale information of the camouflage target;

[0098] Step 2.2: Change the disguised identification target and repeat step 2.1 to obtain the RGB three-channel grayscale information of various disguised identification targets. The RGB three-channel grayscale information of various disguised identification targets is composed of standard RGB channel data.

[0099] Step 2.3: Store the standard RGB channel data in the RGB color database 251, and configure the RGB color database 251 into the color recognition module 241.

[0100] Step 3: Place the calibration whiteboard 141 within the shooting area of ​​the camouflage identification target. The image acquisition unit 101 acquires the spectral image of the calibration whiteboard 141, and the environmental spectrum of the calibration whiteboard is obtained through the spectral reconstruction module 211. After a single calibration, it can work continuously for a period of time until the ambient light color temperature changes significantly, and then the next calibration is performed.

[0101] Step 4: Return the calibration whiteboard 141 to the calibration whiteboard storage fixture 142. The image acquisition unit 101 acquires the grayscale image of the camouflage identification target shooting area. The spectral reconstruction module 211 converts the grayscale image into spectral data cubes.

[0102] Step 5: The true color restoration module 221 converts the spectral data cubically into an RGB image, specifically as follows:

[0103] Step 5.1: Multiply the spectral dimension matrix of each pixel in the spectral data cube obtained in Step 4 by the CIE-RGB1931 chromaticity coordinate curve to obtain the RGB channel grayscale value of each pixel:

[0104] RGB(x,y) = CIE*S(x,y)

[0105] Where S(x,y) represents the spectral dimension matrix of the pixel at spatial coordinate (x,y), RGB(x,y) represents the RGB channel grayscale value of the pixel at spatial coordinate (x,y), and CIE represents the CIE-RGB1931 chromaticity coordinate curve.

[0106] Step 5.2: Arrange the gray values ​​of the RGB channels of each pixel in the order of the cubic space coordinates of the spectral data to obtain the RGB image.

[0107] Step 6: The color recognition module 241 selects the corresponding standard RGB channel data from the RGB color database 251 configured in Step 2 based on the camouflage recognition target, and performs preliminary color filtering on the RGB image obtained in Step 5 based on the standard RGB channel data to obtain the camouflage recognition pixel area with the color of the camouflage recognition target. The specific method is as follows:

[0108] Step 6.1: The color recognition module 241 selects the corresponding standard RGB channel data from the RGB color database 251 configured in step 2 according to the disguised recognition target. It normalizes the standard RGB channel data using the grayscale value of channel G in the standard RGB channel data as the standard to obtain normalized standard RGB channel data. The normalized standard RGB channel data includes standard R channel grayscale data, standard G channel grayscale data and standard B channel grayscale data.

[0109] The specific method of normalization is as follows:

[0110] A. Divide the grayscale value of channel G in the standard RGB channel data by 1 to obtain the coefficient a;

[0111] B. Divide the gray values ​​r1 of channel R, g1 of channel G, and b1 of channel B in the standard RGB channel data by the coefficient a to obtain the normalized channel gray values ​​r, g, and b respectively.

[0112] C. Integrate r, g, and b into normalized standard RGB channel data.

[0113] Step 6.2: Extract the RGB channel data of each pixel in the RGB image obtained in Step 5. According to the normalization method in Step 6.1, normalize the RGB channel data using the gray value of channel G in the RGB channel data of each pixel as the standard to obtain the normalized RGB channel data of each pixel. The normalized RGB channel data includes the gray value of R channel, G channel and B channel.

[0114] Step 6.3: Calculate the Euclidean distance between the normalized RGB channel data and the normalized standard RGB channel data for each pixel using the following formula:

[0115]

[0116] Where L represents the Euclidean distance between the normalized RGB channel data and the normalized standard RGB channel data, the smaller the Euclidean distance, the smaller the difference between the two sets of data;

[0117] r0, g0, and b0 represent standard R-channel grayscale data, standard G-channel grayscale data, and standard B-channel grayscale data, respectively.

[0118] r, g, and b represent the R-channel grayscale data, G-channel grayscale data, and B-channel grayscale data of pixels in an RGB image, respectively.

[0119] Step 6.4: Set a color recognition threshold. Compare the Euclidean distance L between the normalized RGB channel data and the normalized standard RGB channel data of each pixel with the color recognition threshold. Pixels whose normalized RGB channel data L is less than the color recognition threshold are recorded as target pixels. The area formed by the target pixels is the camouflage recognition pixel area with the ability to camouflage the target color. The color recognition threshold should not be too small; it only needs to achieve basic color differentiation.

[0120] Step 7: Select the spectral data cube corresponding to the camouflage recognition pixel region from the spectral data cube obtained in Step 4. The spectral correction module 231 uses the environmental spectrum obtained in Step 3 to correct the spectral data cube corresponding to the camouflage recognition pixel region, obtaining the corrected spectrum of each pixel within the camouflage recognition pixel region. The specific method for correcting the spectral data cube is as follows:

[0121] Divide the cube of the spectral data corresponding to the camouflage-identified pixel region by the environmental spectrum one by one.

[0122] Step 8: The spectral classification and recognition module 261 selects the corresponding camouflage object reflectance spectrum and camouflage material reflectance spectrum from the spectral reflectance database 271 configured in Step 1 according to the camouflage recognition target. It then compares the corrected spectrum of each pixel in the camouflage recognition pixel area with the camouflage object reflectance spectrum and the camouflage material reflectance spectrum. If a pixel contains the spectral features of the camouflage recognition target, the pixel is selected; otherwise, the comparison continues to the next pixel. The area composed of pixels containing the spectral features of the camouflage recognition target in the camouflage recognition pixel area is recorded as the camouflage pixel area.

[0123] The specific method for step 8 is as follows:

[0124] Step 8.1: The spectral classification and recognition module 261 selects the corresponding camouflage object reflectance spectrum and camouflage material reflectance spectrum from the spectral reflectance database 271 configured in step 1 according to the camouflage recognition target.

[0125] Step 8.2: Select the characteristic spectral band of the corrected spectrum of each pixel in the camouflage recognition pixel area, and normalize the spectrum of the corresponding pixel using the characteristic spectral band as the standard to obtain the normalized spectrum of each pixel in the camouflage recognition pixel area.

[0126] Select the characteristic spectral bands of the reflectance spectrum of the camouflage object and the characteristic spectral bands of the reflectance spectrum of the camouflage material, and normalize the reflectance spectra of the camouflage object and the camouflage material respectively using the characteristic spectral bands as the standard to obtain the normalized reflectance spectra of the camouflage object and the normalized reflectance spectra of the camouflage material.

[0127] Step 8.3: Calculate the Euclidean distance between the normalized spectrum of each pixel in the characteristic spectral band and the normalized reflectance spectrum of the camouflage object, and the Euclidean distance between the normalized spectrum of each pixel and the reflectance spectrum of the camouflage material, to obtain two sets of Euclidean distances for each pixel.

[0128] Step 8.4: Set the spectral recognition threshold. Compare the two sets of Euclidean distances for each pixel obtained in Step 8.3 with the spectral recognition threshold. If both sets of Euclidean distances for a pixel are greater than the spectral recognition threshold, then classify the pixel as "other".

[0129] If at least one set of Euclidean distances of a pixel is less than or equal to the spectral recognition threshold, and the Euclidean distance between the normalized spectrum of the pixel and the reflectance spectrum of the camouflage object is less than the Euclidean distance between the normalized spectrum and the reflectance spectrum of the camouflage material, then the pixel is classified as a camouflage object.

[0130] If at least one set of Euclidean distances of a pixel is less than or equal to the spectral recognition threshold, and the Euclidean distance between the normalized spectrum of the pixel and the reflectance spectrum of the camouflage object is greater than the Euclidean distance between the normalized spectrum and the reflectance spectrum of the camouflage material, then the pixel is classified as a camouflage material.

[0131] Step 8.5: Mark the pixels that are classified as camouflage materials. The area composed of such pixels is the camouflage pixel area.

[0132] Step 9: Mark the spoofed pixel areas in the RGB image obtained in step 5 to obtain a color image with spoofed markings, and send it to the display storage unit 301 through the recognition result output module 281. The display storage unit 301 displays and stores the color image with spoofed markings to complete the spoofing recognition.

Claims

1. A multispectral camouflage recognition imaging system characterized by: The calibration whiteboard (141), the image acquisition unit (101) and the data processing unit (201) are electrically connected, and the display storage unit (301) is electrically connected with the data processing unit (201); The calibration whiteboard (141) is located in the shooting area of the camouflage identification target during calibration; The image acquisition unit (101) comprises a band-pass filter (111), an imaging lens group (121) and a snapshot spectral imaging chip (131) arranged in sequence along the light path of the camouflage identification target, and a sensor circuit group (132) connected with the snapshot spectral imaging chip (131); The data processing unit (201) comprises a spectral reconstruction module (211), a true color restoration module (221), a color recognition module (241), a spectral classification and identification module (261) and an identification result output module (281) connected in sequence, and a spectral correction module (231); The input end of the spectral correction module (231) is connected with the output end of the spectral reconstruction module (211), and the output end is connected with the input end of the spectral classification and identification module (261); The snapshot spectral imaging chip (131) is electrically connected with the spectral reconstruction module (211) through the sensor circuit group (132), and the identification result output module (281) is electrically connected with the display storage unit (301); The CIE-RGB1931 chromaticity curve is configured in the true color restoration module (221); The RGB color database (251) is configured in the color recognition module (241), and the RGB color database (251) comprises standard RGB channel data of multiple camouflage identification targets; The spectral reflectivity database (271) is configured in the spectral classification and identification module (261), and the spectral reflectivity database (271) comprises camouflage object reflectivity spectra of multiple approximate target camouflage objects and camouflage material reflectivity spectra of multiple approximate target camouflage materials, which are obtained by the following steps: Step A.1: Obtain the linear difference of the characteristic spectral segment of the approximate target camouflage object and the approximate target camouflage material, and prepare a certain number of approximate target camouflage objects and approximate target camouflage materials as samples for constructing the spectral reflectivity database (271); Step A.2: Place the calibration whiteboard (141) in sunlight, and the image acquisition unit (101) acquires the spectral image of the calibration whiteboard (141), and obtains the sunlight environment spectrum by using the spectral reconstruction module (211); Step A.3: The image acquisition unit (101) respectively acquires the spectral images of the approximate target camouflage object and the approximate target camouflage material, and obtains the camouflage object spectrum and the camouflage material spectrum by using the spectral reconstruction module (211); Step A.4: The spectral correction module (231) corrects the camouflage object spectrum and the camouflage material spectrum obtained in step A.3 according to the sunlight environment spectrum obtained in step A.2, to obtain the camouflage object reflectivity spectrum and the camouflage material reflectivity spectrum, and selects the characteristic spectral segment of the camouflage object reflectivity spectrum and the camouflage material reflectivity spectrum and records them. Step A.5, replace the approximate target camouflage object and the approximate target camouflage material, repeat step A.3-step A.4, obtain the characteristic spectrum of the multiple sets of camouflage object reflectance spectrum and camouflage material reflectance spectrum of all the approximate target camouflage objects and approximate target camouflage materials prepared in step A.1; Step A.6, determine whether the characteristic spectrum of the multiple sets of camouflage object reflectance spectrum and camouflage material reflectance spectrum obtained in step A.5 exists the linear difference rule of the characteristic spectrum obtained in step A.1, thereby verifying its effectiveness, and obtaining the characteristic spectrum of the effective camouflage object reflectance spectrum and camouflage material reflectance spectrum; Step A.7, store the characteristic spectrum of the effective camouflage object reflectance spectrum and camouflage material reflectance spectrum as the standard spectrum in the spectral reflectance database (271), and configure the spectral reflectance database (271) into the spectral classification and identification module (261).

2. The multispectral camouflage recognition imaging system of claim 1, wherein: The data processing unit (201) is an external computer or an internal FPGA.

3. The multispectral camouflage recognition imaging system according to claim 1 or 2, characterized in that: The display storage unit (301) includes a digital image display device (311) and a data storage device (321); The digital image display device (311) and the data storage device (321) are respectively electrically connected with the identification result output module (281).

4. A multispectral camouflage recognition method based on the multispectral camouflage recognition imaging system of any one of claims 1-3, characterized in that, Comprising the following steps: Step 1, obtain the spectral reflectance database (271) including the standard spectrum, and configure it into the spectral classification and identification module (261); the standard spectrum includes the camouflage object reflectance spectrum of multiple approximate target camouflage objects and the camouflage material reflectance spectrum of multiple approximate target camouflage materials; Step 2, obtain the RGB color database (251) including the standard RGB channel data, and configure it into the color identification module (241); the RGB color database (251) includes the standard RGB channel data of multiple camouflage identification targets; Step 3, place the calibration whiteboard (141) into the shooting area of the camouflage identification target, the image acquisition unit (101) acquires the spectral image of the calibration whiteboard (141), and the spectral reconstruction module (211) obtains the environmental spectrum of the calibration whiteboard (141); Step 4, remove the calibration whiteboard (141), the image acquisition unit (101) acquires the gray scale image of the shooting area of the camouflage identification target, and the spectral reconstruction module (211) converts the gray scale image into a spectral data cube; Step 5, the true color restoration module (221) converts the spectral data cube into an RGB image; Step 6, the color identification module (241) selects the corresponding standard RGB channel data of the camouflage identification target in the RGB color database (251) configured in step 2 according to the camouflage identification target, and preliminarily color screens the RGB image obtained in step 5 according to the standard RGB channel data, to obtain a camouflage identification pixel region with the color of the camouflage identification target. Step 7, select the spectral data cube corresponding to the camouflage identification pixel region from the spectral data cube obtained in step 4, and correct the spectral data cube corresponding to the camouflage identification pixel region by using the ambient spectrum obtained in step 3 by the spectral correction module (231), to obtain the corrected spectrum of each pixel in the camouflage identification pixel region; Step 8, according to the camouflage identification target, select the corresponding camouflage object reflectivity spectrum and camouflage material reflectivity spectrum in the spectral reflectivity database (271) configured in step 1 by the spectral classification and identification module (261), and compare the corrected spectrum of each pixel in the camouflage identification pixel region with the camouflage object reflectivity spectrum and the camouflage material reflectivity spectrum respectively, if the pixel contains the spectral characteristics of the camouflage identification target, select the pixel, otherwise continue to compare the next pixel; the region composed of the pixels containing the spectral characteristics of the camouflage identification target in the camouflage identification pixel region is recorded as the camouflage pixel region; Step 9, mark the camouflage pixel region in the RGB image obtained in step 5 to obtain a color image with camouflage marking, and send it to the display storage unit (301) through the identification result output module (281), and the display storage unit (301) displays and stores the color image with camouflage marking, and completes the camouflage identification.

5. The multispectral camouflage recognition method of claim 4, wherein, The step 8 is specifically: Step 8.1, according to the camouflage identification target, select the corresponding camouflage object reflectivity spectrum and camouflage material reflectivity spectrum in the spectral reflectivity database (271) configured in step 1 by the spectral classification and identification module (261); Step 8.2, select the characteristic spectral band of the corrected spectrum of each pixel in the camouflage identification pixel region, and normalize the spectrum of the corresponding pixel by taking the characteristic spectral band as the standard, to obtain the normalized spectrum of each pixel in the camouflage identification pixel region; select the characteristic spectral band of the characteristic spectral band of the camouflage object reflectivity spectrum and the characteristic spectral band of the camouflage material reflectivity spectrum, and normalize the camouflage object reflectivity spectrum and the camouflage material reflectivity spectrum respectively by taking the characteristic spectral band as the standard, to obtain the normalized camouflage object reflectivity spectrum and the normalized camouflage material reflectivity spectrum; Step 8.3, calculate the Euclidean distance between the normalized spectrum of each pixel in the characteristic spectral band and the normalized camouflage object reflectivity spectrum, and the Euclidean distance between the normalized spectrum of each pixel and the camouflage material reflectivity spectrum, to obtain two groups of Euclidean distances of each pixel; Step 8.4, set a spectral identification threshold, compare the two groups of Euclidean distances of each pixel obtained in step 8.3 with the spectral identification threshold respectively, if the two groups of Euclidean distances of a pixel are both greater than the spectral identification threshold, the pixel is classified as other; if at least one group of Euclidean distances of the pixel is less than or equal to the spectral identification threshold, and the Euclidean distance between the normalized spectrum of the pixel and the camouflage object reflectivity spectrum is less than the Euclidean distance between the normalized spectrum and the camouflage material reflectivity spectrum, the pixel is classified as a camouflage object. If the Euclidean distance of at least one group of pixels is less than or equal to the spectral recognition threshold, and the Euclidean distance between the normalized spectrum of the pixel and the reflectivity spectrum of the camouflage object is greater than the Euclidean distance between the normalized spectrum and the reflectivity spectrum of the camouflage material, the pixel is classified as a camouflage material; Step 8.5, mark the pixels classified as camouflage materials, and the area composed of such pixels is a camouflage pixel area.

6. The multispectral camouflage recognition method according to claim 4 or 5, characterized in that, The step 6 is specifically: Step 6.1, the color recognition module (241) selects the corresponding standard RGB channel data in the RGB color database (251) configured in step 2 according to the camouflage recognition target, and normalizes the standard RGB channel data with the gray value of channel G in the standard RGB channel data as the standard to obtain normalized standard RGB channel data, wherein the normalized standard RGB channel data includes standard R channel gray data, standard G channel gray data and standard B channel gray data; Step 6.2, extract the RGB channel data of each pixel in the RGB image obtained in step 5, and normalize the RGB channel data of each pixel according to the normalization method of step 6.1 with the gray value of channel G in the RGB channel data of each pixel as the standard to obtain the normalized RGB channel data of each pixel, wherein the normalized RGB channel data includes R channel gray data, G channel gray data and B channel gray data; Step 6.3, calculate the Euclidean distance L between the normalized RGB channel data of each pixel and the normalized standard RGB channel data according to the following formula: Wherein, L represents the Euclidean distance between the normalized RGB channel data and the normalized standard RGB channel data, and the smaller the Euclidean distance, the smaller the difference between the two groups of data; r0, g0, b0 represent the standard R channel gray data, the standard G channel gray data and the standard B channel gray data respectively; r, g, b represent the R channel gray data, the G channel gray data and the B channel gray data of the pixel in the RGB image respectively; Step 6.4, set a color recognition threshold, compare the Euclidean distance L between the normalized RGB channel data of each pixel and the normalized standard RGB channel data with the color recognition threshold, mark the pixel corresponding to the normalized RGB channel data with L less than the color recognition threshold as a target pixel, and the area composed of the target pixels is a camouflage recognition pixel area with the color of the camouflage recognition target.

7. The multispectral camouflage recognition method of claim 6, wherein, In step 6.1, the specific method of normalization is: A, divide the gray value of channel G in the standard RGB channel data by 1 to obtain a coefficient a; B, divide the gray values r1 of channel R, g1 of channel G and b1 of channel B in the standard RGB channel data by the coefficient a to obtain the normalized channel gray values r, g and b respectively; C, integrate r, g and b into the normalized standard RGB channel data.

8. The multispectral camouflage recognition method of claim 7, wherein, The step 5 is specifically: Step 5.1, multiply the spectral dimension matrix of each pixel in the spectral data cube obtained in step 4 by the CIE-RGB1931 chromaticity coordinate curve to obtain the RGB channel gray value of each pixel: RGB(x, y) = CIE * S(x, y) Wherein S(x, y) represents the spectral dimension matrix of the pixel at the spatial coordinates (x, y), RGB(x, y) represents the RGB channel grayscale value of the pixel at the spatial coordinates (x, y), and CIE represents the CIE-RGB1931 chromaticity coordinate curve. Step 5.2, arranging the RGB channel grayscale values of each pixel in the order of the spectral data cubic space coordinates to obtain an RGB image.

9. The multispectral camouflage recognition method of claim 8, wherein, The step 2 is specifically: Step 2.1, placing the camouflage recognition target under sunlight and recording the RGB three-channel grayscale information of the camouflage recognition target by using the achromatic RGB color camera; Step 2.2, replacing the camouflage recognition target, repeating step 2.1, and obtaining the RGB three-channel grayscale information of multiple camouflage recognition targets, wherein the RGB three-channel grayscale information of the multiple camouflage recognition targets constitutes standard RGB channel data; Step 2.3, storing the standard RGB channel data in the RGB color database (251) and configuring the RGB color database (251) into the color recognition module (241).

Images