Detection and identification method, device and equipment based on target space-time spectrum feature modeling

Abstract

The invention provides a detection and identification method, device and equipment based on target space-time spectrum feature modeling. The detection and identification method based on target space-time spectrum feature modeling comprises the following steps: acquiring a to-be-identified image sequence; performing image noise reduction processing on the to-be-recognized image sequence to obtain a to-be-recognized noise-reduced image sequence; generating target alternative screening points through the to-be-recognized noise reduction image sequence; and based on the target alternative screening points and the target feature database, performing data matching processing by adopting a data matching network to obtain a target recognition result. The construction of the target feature database not only considers the infrared radiation characteristic of the target, but also considers the imaging effect of the detector, so that the refinement degree and integrity of the target feature database are improved, and the accuracy of target recognition is improved on the basis; in addition, the target feature database contains target features under various scenes and conditions, so that the method has higher universality, and the adaptability of different detectors and systems is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of signal processing and pattern recognition, and in particular to a detection and recognition method, device and equipment based on target spatiotemporal spectral feature modeling. Background Art

[0002] With the continuous development of space target detection and recognition technology, algorithms based on deep learning have been widely used in the analysis of space-based detector image sequences. This type of algorithm can accurately detect and identify targets from images through complex calculation processes, but it has high hardware requirements and a long processing flow, and is more suitable for the early warning stage. However, in interceptor applications, due to hardware limitations and real-time requirements, it is difficult to directly adopt the same target detection and recognition algorithm as the early warning stage. Therefore, how to achieve efficient and accurate target detection and recognition in the interceptor scenario has become a technical problem that needs to be solved in the current field of space target recognition.

[0003] In the existing technology, there have been some related studies on the problems of space target characteristic modeling and target detection and recognition. For example, some scholars have constructed a fine geometric model of space targets through methods such as texture addition, mesh generation and material setting, and combined with the thermal mesh model and infrared radiation transmission link model to realize the full-link imaging simulation of space-based infrared cameras. In terms of target detection and recognition, other scholars have proposed a time series classification algorithm based on sparse representation, combined with the idea of ​​deep learning, to further optimize the detection performance of infrared point targets under complex backgrounds. These studies provide important theoretical basis and technical support for space target characteristic modeling and detection and recognition.

[0004] However, the existing technology still has some shortcomings. First, there are few studies on modeling and simulating the comprehensive analysis of the infrared radiation characteristics of ballistic targets and the entire imaging effects of the detection system to establish a refined database, resulting in insufficient integrity and refinement of the target feature database. Secondly, the existing target detection and recognition algorithms are mostly designed for specific systems, lack universal extensions, and are difficult to adapt to different scenarios and application requirements. In addition, the existing algorithms have limitations in real-time and hardware adaptability, and are difficult to meet application scenarios with high real-time requirements such as interceptors. These problems limit the further development and application of space target detection and recognition technology. Summary of the invention

[0005] In order to solve the above problems existing in the prior art, the present invention provides a detection and identification method, device and equipment based on target spatiotemporal feature modeling.

[0006] The technical problem to be solved by the present invention is achieved through the following technical solutions:

[0007] In a first aspect, the present invention provides a detection and recognition method based on target spatiotemporal spectral feature modeling, comprising:

[0008] Obtain the image sequence to be recognized; the image sequence to be recognized is the image sequence captured and transmitted back by the detector;

[0009] Perform image noise reduction processing on the image sequence to be recognized to obtain the denoised image sequence to be recognized;

[0010] Generate target alternative screening points through the denoised image sequence to be recognized;

[0011] Based on the target alternative screening points and the target feature database, use a data matching network to perform data matching processing to obtain the target recognition result;

[0012] The target feature database is obtained through modeling and simulation using the target surface grid model and multiple groups of simulation parameters; the multiple groups of simulation parameters are obtained by traversing the simulation data set; the simulation data set includes: detector band information and simulated target detection information; the simulated target detection information includes: simulated target surface temperature field distribution, simulated target surface material characteristic parameters, simulated target data affected by external radiation sources, simulated target motion data, detector observation direction and detector observation position.

[0013] Optionally, the modeling and simulation process of the target feature database includes:

[0014] Obtain the simulation data set; among them, the simulated target detection information is obtained based on a preset step size and a traversal processing method;

[0015] Perform a loop traversal on the simulated target detection information to generate multiple groups of simulation parameters;

[0016] Traverse the detector band information in sequence. Under the currently obtained detector band information during traversal, substitute the multiple groups of simulation parameters into the target surface grid model for modeling and simulation processing in sequence to obtain the initial target feature data set;

[0017] Construct the target feature database from the initial target feature data sets corresponding to all detector band information.

[0018] Optionally, traverse the detector band information in sequence. Under the currently obtained detector band information during traversal, substitute the multiple groups of simulation parameters into the target surface grid model for modeling and simulation processing in sequence to obtain the initial target feature data set, including:

[0019] Traverse the detector band information in sequence to obtain the current detector band information;

[0020] In the target surface grid model, split the simulated target surface into multiple surface elements;

[0021] Calculate the image plane coordinates of the multiple surface elements through the object-image conversion algorithm;

[0022] Based on the simulation parameters and the current detector band information, using the infrared radiation theory, calculate the intrinsic radiation and reflected radiation of multiple surface elements;

[0023] Based on the image plane coordinates, intrinsic radiation and reflected radiation of multiple surface elements, calculate the radiation flux using the radiation flux calculation method; the radiation flux is the radiation flux of multiple surface elements passing through the detector entrance pupil;

[0024] Utilize multiple surface elements and the radiation flux to add the imaging effect of the optical system to obtain the second grid conversion result;

[0025] Perform image plane coordinate conversion processing on the second grid conversion result to obtain the initial target feature dataset.

[0026] Optionally, utilize multiple surface elements and the radiation flux to add the imaging effect of the optical system to obtain the second grid conversion result, including:

[0027] Perform fine grid splitting processing on multiple surface elements to obtain refined grids;

[0028] Convert the image plane coordinates of multiple surface elements into the refined grids to obtain the first grid conversion result;

[0029] Combine the radiation flux to add the imaging effect of the optical system to the first grid conversion result to obtain the second grid conversion result.

[0030] Optionally, perform image plane coordinate conversion processing on the second grid conversion result to obtain the initial target feature dataset, including:

[0031] Perform fine grid removal on the second grid conversion result through downsampling processing, and return the result of fine grid removal to the image plane coordinates of the detector to obtain the initial target feature dataset.

[0032] Optionally, perform image noise reduction processing on the image sequence to be recognized to obtain the image sequence to be recognized after noise reduction, including:

[0033] Perform image noise reduction processing on the image sequence to be recognized using the weighted mean processing method to obtain the image sequence to be recognized after noise reduction.

[0034] Optionally, generate target alternative screening points from the image sequence to be recognized after noise reduction, including:

[0035] Process the image sequence to be recognized after noise reduction using the spatial local contrast and temporal local contrast methods to obtain target alternative points; the target alternative points contain the position, pixel distribution, and gray value information of each target alternative point;

[0036] Perform inversion calculation on the target alternative points to obtain the target alternative screening points.

[0037] Optionally, based on the target alternative screening points and the target feature database, a data matching network is used to perform data matching processing to obtain a target recognition result, including:

[0038] Obtain the matching range corresponding to the target alternative screening points;

[0039] Use the data in the target feature database within the matching range as the first target matching database;

[0040] Perform gray quantization and downsampling operations on the first target matching database in sequence to obtain a second target matching database;

[0041] Use the data matching network to perform data matching processing on the second target matching database and the target alternative screening points to obtain a target recognition result.

[0042] In a second aspect, the present invention provides a detection and recognition device based on target spatio-temporal spectrum feature modeling. The detection and recognition device based on target spatio-temporal spectrum feature modeling includes: an acquisition unit, a preprocessing unit, a screening unit, and a data matching unit;

[0043] The acquisition unit is used to: acquire an image sequence to be recognized; the image sequence to be recognized is an image sequence captured and transmitted back by a detector;

[0044] The preprocessing unit is used to: perform image noise reduction processing on the image sequence to be recognized to obtain a denoised image sequence to be recognized;

[0045] The screening unit is used to: generate target alternative screening points through the denoised image sequence to be recognized;

[0046] The data matching unit is used to: based on the target alternative screening points and the target feature database, use a data matching network to perform data matching processing to obtain a target recognition result;

[0047] The target feature database is obtained by performing modeling and simulation using a target surface grid model and multiple sets of simulation parameters; the multiple sets of simulation parameters are obtained by traversing a simulation data set; the simulation data set includes: detector band information and simulated target detection information; the simulated target detection information includes: simulated target surface temperature field distribution, simulated target surface material characteristic parameters, simulated target data affected by external radiation sources, simulated target motion data, detector observation direction and detector observation position.

[0048] In a third aspect, the present invention provides a detection and recognition device based on target spatio-temporal spectrum feature modeling, including: a processor, a storage medium, and a bus. The storage medium stores machine-readable instructions executable by the processor. When the detection and recognition device based on target spatio-temporal spectrum feature modeling runs, the processor communicates with the storage medium through the bus, and the processor executes the machine-readable instructions to perform the steps of the detection and recognition method based on target spatio-temporal spectrum feature modeling according to any one of the above first aspects.

[0049] The present invention provides a detection and recognition method, device, and equipment based on target spatio-temporal spectrum feature modeling. Among them, a detection and recognition method based on target spatio-temporal spectrum feature modeling includes: obtaining an image sequence to be recognized; the image sequence to be recognized is an image sequence captured and transmitted back by a detector; performing image noise reduction processing on the image sequence to be recognized to obtain a denoised image sequence to be recognized; generating target alternative screening points through the denoised image sequence to be recognized; based on the target alternative screening points and a target feature database, performing data matching processing using a data matching network to obtain a target recognition result; the target feature database is obtained by performing modeling and simulation using a target surface grid model and multiple groups of simulation parameters; the multiple groups of simulation parameters are obtained by traversing a simulation data set; the simulation data set includes: detector band information and simulated target detection information; the simulated target detection information includes: simulated target surface temperature field distribution, simulated target surface material characteristic parameters, simulated target data affected by external radiation sources, simulated target motion data, detector observation direction, and detector observation position. In the present invention, a refined target feature database for the entire ballistic target link is formed through the target surface grid model and multiple groups of simulation parameters. Since the construction of the target feature database not only considers the infrared radiation characteristics of the target but also takes into account the imaging effect of the detector, the refinement degree and integrity of the target feature database are improved, and on this basis, the accuracy of target recognition is improved; in addition, the target feature database contains target features under various scenarios and conditions, so it has stronger universality and improves the adaptability to different detectors and systems; finally, by using a data matching network to perform data matching processing between the imaging feature database and the target alternative screening points, the speed of target recognition and the flexibility in hardware adaptation can be greatly improved.

[0050] The following will further describe the present invention in detail with reference to the accompanying drawings and embodiments. Description of the Drawings

[0051] Figure 1 It is a schematic flowchart of a detection and recognition method based on target spatio-temporal spectrum feature modeling provided by an embodiment of the present invention;

[0052] Figure 2 Exemplarily shows a flowchart block diagram of a detection and recognition method based on target spatio-temporal spectrum feature modeling;

[0053] Figure 3 The structural schematic diagram of a detection and recognition device based on target spatio-temporal spectrum feature modeling provided by an embodiment of the present invention;

[0054] Figure 4 The structural schematic diagram of a detection and recognition device based on target spatio-temporal spectrum feature modeling provided by an embodiment of the present invention. Specific embodiments

[0055] The present invention will be further described in detail below in conjunction with specific embodiments, but the implementation manners of the present invention are not limited thereto.

[0056] In order to improve the accuracy of target recognition, the speed of target recognition, and the flexibility in terms of hardware adaptation, an embodiment of the present invention provides a detection and recognition method based on target spatio-temporal spectrum feature modeling. Figure 1 The flowchart of a detection and recognition method based on target spatio-temporal spectrum feature modeling provided by an embodiment of the present invention. As Figure 1 shown, it includes:

[0057] S101. Obtain an image sequence to be recognized.

[0058] In an embodiment of the present invention, the image sequence to be recognized is an image sequence captured and transmitted back by a detector. The image sequence to be recognized not only records the dynamic behaviors of the target and the accompanying object during the inertial orbit flight, but also contains rich information for subsequent analysis and recognition. Specifically, the image sequence to be recognized contains the following key information: spatial position information, infrared radiation intensity information, spectral feature information, time variation information, background information, and noise and interference information.

[0059] S102. Perform image denoising processing on the image sequence to be recognized to obtain a denoised image sequence to be recognized.

[0060] Optionally, S102 may specifically include:

[0061] Perform image denoising processing on the image sequence to be recognized by using a weighted mean processing method to obtain a denoised image sequence to be recognized.

[0062] Specifically, in this implementation manner, the superposition window length is determined according to the actual situation (the target should basically not generate pixel offset within this window length), and the randomness of the noise distribution and energy is reduced by superposing and taking the weighted mean. At the same time, the energy of the target area is strengthened to a certain extent for subsequent recognition, and finally a denoised image sequence to be recognized is obtained.

[0063] S103. Generate target alternative screening points through the denoised image sequence to be recognized.

[0064] Optionally, S103 may specifically include:

[0065] The method of spatial local contrast and temporal local contrast is used to process the image sequence of the noise-reduced image to be recognized, and target candidate points are obtained; the target candidate points include the positions, pixel distributions, and gray value information of each target candidate point.

[0066] Inverse calculation is performed on the target candidate points to obtain target candidate screening points.

[0067] Specifically, the calculation process of the spatial local contrast is as follows:

[0068] Taking the current pixel of the image in the image sequence of the noise-reduced image to be recognized as the center, a small neighborhood of the current pixel is taken to form a central block, and the small neighborhood pixels in eight directions outward from the central block are selected as neighborhood pixels. The weighted average of the neighborhood pixels is calculated as the gray value representative value of the whole neighborhood pixels, denoted as SB(i, j, k). Then, the spatial domain local contrast result SLCM(i, j, k) is:

[0069]

[0070] Among them, F(i, j, k) represents the true gray value of the k-th frame image at the current pixel (i, j), i represents the horizontal pixel value of the current pixel, and j represents the vertical pixel value of the current pixel.

[0071] The calculation process of the temporal local contrast is as follows:

[0072] The current frame image in the image sequence of the noise-reduced image to be recognized is used to subtract the images separated by m frames before and after respectively to obtain two difference images, and the two difference images are multiplied to obtain the temporal local contrast TLCM(i, j, n). After obtaining the spatial domain local contrast SLCM(i, j, k) and the temporal local contrast TLCM(i, j, n), the spatio-temporal local contrast is obtained by using the spatial domain local contrast and the temporal local contrast. The spatio-temporal local contrast STLCM(i, j, n) is:

[0073]

[0074] Due to the point spread effect, the target energy may spread to its surrounding pixels. Therefore, in the calculation result, if the spatio-temporal local contrast of the current pixel (i, j) is lower than the gray value of a certain pixel in its eight-neighborhood, it is considered that the current pixel (i, j) is not a target point. All target candidate points are finally obtained by using the above processing process.

[0075] After obtaining the target alternative points, starting from the target imaging features, data such as the radiant flux of the target entering the detector and the irradiance reaching the detector are calculated through inverse photoelectric conversion and image gray scale, and further the inversion of the radiation and temperature characteristic data of the target alternative points is completed, and the approximate radiation characteristic data are initially obtained. Then, combined with the characteristics that the real target should have, some target alternative points with large differences from the radiation characteristics of the real target are excluded, which plays a certain primary screening role for the subsequent database comparison and matching, can narrow the matching range, and improve the matching efficiency.

[0076] S104. Based on the target alternative screening points and the target feature database, use a data matching network to perform data matching processing to obtain the target recognition result.

[0077] The target feature database is obtained by performing modeling and simulation using the target surface grid model and multiple groups of simulation parameters; the multiple groups of simulation parameters are obtained by traversing the simulation data set; the simulation data set includes: detector band information and simulated target detection information; the simulated target detection information includes: simulated target surface temperature field distribution, simulated target surface material characteristic parameters, simulated target data affected by external radiation sources, simulated target motion data, detector observation direction and detector observation position.

[0078] To fully illustrate the detection and recognition method based on target spatio-temporal spectrum feature modeling provided by the embodiments of the present invention, Figure 2 exemplarily shows a flowchart of a detection and recognition method based on target spatio-temporal spectrum feature modeling. As Figure 2 shown, first, an image sequence to be recognized is obtained, and image noise reduction processing is performed on the basis of the image sequence to be recognized to obtain a noise-reduced image sequence to be recognized. Then, preliminary screening of target points is performed according to radiation characteristics to obtain target alternative screening points. Before this, the target feature database can be obtained first. Specifically, first, multiple groups of simulation parameters are constructed through the simulation data set. Then, using the target surface grid model and multiple groups of simulation parameters, the target feature database is obtained through modeling and simulation. Next, the matching range screening, gray scale quantization, and downsampling processing are performed on the target feature database to obtain the second target matching database. On the basis of the second target matching database and the target alternative screening points, use a data matching network to perform data matching processing on the second target matching database and the target alternative screening points, and determine whether the matching result meets the matching threshold. If the matching threshold is met, select the current target alternative screening point as the recognized target. If the screening threshold is not met, exclude the current target alternative screening point, and thus obtain all the final target recognition results.

[0079] An embodiment of the present invention provides a detection and recognition method based on target spatio-temporal spectrum feature modeling. A refined target feature database for the entire ballistic target link is formed through a target surface grid model and multiple sets of simulation parameters. Since the construction of the target feature database not only considers the infrared radiation characteristics of the target but also takes into account the imaging effect of the detector, the refinement degree and integrity of the target feature database are improved, and on this basis, the accuracy of target recognition is enhanced. In addition, the target feature database contains target features under various scenarios and conditions, so it has stronger universality and improves the adaptability to different detectors and systems. Finally, by using a data matching network to perform data matching processing on the imaging feature database and target alternative screening points, the speed of target recognition and the flexibility in terms of hardware adaptation can be greatly improved.

[0080] Optionally, S104 may specifically include:

[0081] Obtain the matching range corresponding to the target alternative screening points;

[0082] Use the data in the target feature database within the matching range as the first target matching database;

[0083] Perform grayscale quantization and downsampling operations on the first target matching database in sequence to obtain a second target matching database;

[0084] Use the data matching network to perform data matching processing on the second target matching database and the target alternative screening points to obtain the target recognition result.

[0085] It can be understood that performing certain quantization and downsampling processing on the data within the matching range in the target feature database (the first target matching database) before data matching processing can make the highly refined data in the first target matching database closer to the actual imaging result.

[0086] The data matching network can use a pre-trained data matching network in the prior art or can obtain the data matching network by fine-tuning the initial data matching network.

[0087] Optionally, the modeling and simulation process of the target feature database includes:

[0088] Obtain a simulation data set; wherein the simulated target detection information is obtained based on a preset step size and a traversal processing method;

[0089] Perform a loop traversal on the simulated target detection information to generate multiple sets of simulation parameters;

[0090] Traverse the detector band information in sequence. Under the currently obtained detector band information during traversal, substitute the multiple sets of simulation parameters into the target surface grid model for modeling and simulation processing to obtain an initial target feature data set;

[0091] Construct a target feature database from the initial target feature datasets corresponding to all detector band information.

[0092] Optionally, traverse the detector band information in sequence. Under the currently obtained detector band information during the traversal, substitute multiple groups of simulation parameters into the target surface grid model for modeling and simulation processing to obtain an initial target feature dataset, including:

[0093] Traverse the detector band information in sequence to obtain the current detector band information;

[0094] Split the simulated target surface into multiple surface elements in the target surface grid model;

[0095] Calculate the image plane coordinates of multiple surface elements through the object-image conversion algorithm;

[0096] Based on the simulation parameters and the current detector band information, adopt the infrared radiation theory to calculate the intrinsic radiation and reflected radiation of multiple surface elements;

[0097] Based on the image plane coordinates, intrinsic radiation, and reflected radiation of multiple surface elements, calculate the radiation flux using the radiation flux calculation method; the radiation flux is the radiation flux of multiple surface elements passing through the detector entrance pupil;

[0098] Add the imaging effect of the optical system using multiple surface elements and the radiation flux to obtain the second grid conversion result;

[0099] Perform image plane coordinate conversion processing on the second grid conversion result to obtain the initial target feature dataset.

[0100] It should be noted that each surface element obtained by splitting contains position, normal direction, and area information. The imaging effect of the added optical system generally includes diffraction and distortion, etc.

[0101] Optionally, add the imaging effect of the optical system using multiple surface elements and the radiation flux to obtain the second grid conversion result, including:

[0102] Perform fine grid splitting processing on multiple surface elements to obtain a refined grid;

[0103] Convert the image plane coordinates of multiple surface elements into the refined grid to obtain the first grid conversion result;

[0104] Combine the radiation flux to add the imaging effect of the optical system to the first grid conversion result to obtain the second grid conversion result.

[0105] Optionally, perform image plane coordinate conversion processing on the second grid conversion result to obtain the initial target feature dataset, including:

[0106] The fine grid is removed from the second grid conversion result through downsampling processing, and the result of the fine grid removal is regressed to the image plane coordinates of the detector to obtain the initial target feature data set.

[0107] The method provided by the embodiments of the present invention can be applied to an electronic device. Specifically, the electronic device can be: a desktop computer, a portable computer, a smart mobile terminal, a server, etc., which are not limited in the embodiments of the present invention.

[0108] Based on the same inventive concept, the embodiments of the present invention also provide a detection and recognition device based on target spatio-temporal spectral feature modeling. Figure 3 As shown in the structural schematic diagram of a detection and recognition device based on target spatio-temporal spectral feature modeling provided by the embodiments of the present invention, Figure 3 it includes: an acquisition unit 301, a preprocessing unit 302, a screening unit 303, and a data matching unit 304;

[0109] The acquisition unit 301 is used to: acquire the image sequence to be recognized;

[0110] The preprocessing unit 302 is used to: perform image noise reduction processing on the image sequence to be recognized to obtain the denoised image sequence to be recognized;

[0111] The screening unit 303 is used to: generate target alternative screening points through the denoised image sequence to be recognized;

[0112] The data matching unit 304 is used to: perform data matching processing based on the target alternative screening points and the target feature database by using a data matching network to obtain the target recognition result;

[0113] The target feature database is obtained by performing modeling and simulation using a target surface grid model and multiple groups of simulation parameters; the multiple groups of simulation parameters are obtained by traversing a simulation data set; the simulation data set includes: detector band information and simulated target detection information; the simulated target detection information includes: simulated target surface temperature field distribution, simulated target surface material characteristic parameters, data on the influence of the simulated target by an external radiation source, simulated target motion data, detector observation direction and detector observation position.

[0114] Figure 4 As shown in the structural schematic diagram of a detection and recognition device based on target spatio-temporal spectral feature modeling provided by the embodiments of the present invention, it includes: a processor 410, a storage medium 420, and a bus 430. The storage medium 420 stores machine-readable instructions executable by the processor 410. When the detection and recognition device based on target spatio-temporal spectral feature modeling runs, the processor 410 communicates with the storage medium 420 through the bus 430, and the processor 410 executes the machine-readable instructions to perform the steps of the above method embodiments. The specific implementation manners and technical effects are similar and will not be elaborated here.

[0115] The storage medium may include a Random Access Memory (RAM), or may also include a Non-Volatile Memory (NVM), such as at least one disk memory. Optionally, the storage medium may also be at least one storage device located far from the aforementioned processor.

[0116] The aforementioned processor may be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it may also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0117] It should be noted that the terms "first", "second", etc. are used to distinguish similar objects, and do not necessarily have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present invention described here can be implemented in an order other than those illustrated or described here. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present invention.

[0118] In the description of this specification, the description with reference to terms such as "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine the different embodiments or examples described in this specification.

[0119] Although the present invention has been described in connection with various embodiments, however, in the process of implementing the claimed invention, those skilled in the art can understand and achieve other variations of the above-described disclosed embodiments by viewing the drawings and the disclosure. In the description of the present invention, the term "comprising" does not exclude other components or steps, the word "a" or "an" does not exclude a plurality of cases, and the meaning of "a plurality" is two or more unless otherwise specifically defined. In addition, certain measures are described in different embodiments, but this does not mean that these measures cannot be combined to produce good results.

[0120] The above content is a further detailed description of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is only limited to these descriptions. For those of ordinary skill in the technical field to which the present invention pertains, without departing from the concept of the present invention, several simple deductions or substitutions can still be made, and all should be regarded as belonging to the protection scope of the present invention.

Claims

1. A detection and recognition method based on target spatiotemporal spectral feature modeling, characterized in that: include: Acquire a sequence of images to be identified; the sequence of images to be identified is a sequence of images captured and transmitted back by the detector; Performing image denoising processing on the image sequence to be identified to obtain a denoised image sequence to be identified; Generate target candidate screening points through the denoised image sequence to be identified; Based on the target candidate screening points and the target feature database, a data matching network is used to perform data matching processing to obtain a target recognition result; The target feature database is obtained by modeling and simulating using a target surface mesh model and multiple sets of simulation parameters; The multiple groups of simulation parameters are obtained by traversing the simulation data set; The simulation data set includes: detector band information and simulated target detection information; the simulated target detection information includes: simulated target surface temperature field distribution, simulated target surface material characteristic parameters, simulated target impact data of external radiation sources, simulated target motion data, detector observation direction and detector observation position.

2. The detection and identification method based on target spatiotemporal spectral feature modeling according to claim 1 is characterized in that: The modeling and simulation process of the target feature database includes: Acquire a simulation data set; wherein the simulated target detection information is obtained based on a preset step size and a traversal processing method; Cycling through the simulated target detection information to generate the multiple sets of simulation parameters; The detector band information is sequentially traversed, and under the current detector band information obtained through the traversal, the multiple groups of simulation parameters are sequentially substituted into the target surface mesh model for modeling and simulation processing to obtain an initial target feature data set; The initial target feature data set corresponding to all detector band information constitutes the target feature database.

3. The detection and identification method based on target spatiotemporal spectral feature modeling according to claim 2 is characterized in that: The step of sequentially traversing the detector band information and sequentially substituting the multiple sets of simulation parameters into the target surface mesh model for modeling and simulation processing under the current detector band information obtained through the traversal to obtain an initial target feature data set includes: Traversing the detector band information in sequence to obtain current detector band information; Splitting the simulated target surface into a plurality of face elements in the target surface mesh model; Calculating the image plane coordinates of the plurality of surface elements by an object-image conversion algorithm; Based on the simulation parameters and the current detector band information, the intrinsic radiation and reflected radiation of the multiple facets are calculated using infrared radiation theory; Based on the image plane coordinates of the multiple facets, the intrinsic radiation and the reflected radiation, a radiation flux is calculated using a radiation flux calculation method; the radiation flux is the radiation flux of the multiple facets passing through the detector entrance pupil; Using the plurality of facets and the radiation flux to add an imaging effect of an optical system, a second grid conversion result is obtained; The second grid transformation result is subjected to image plane coordinate transformation processing to obtain the initial target feature data set.

4. The detection and identification method based on target spatiotemporal spectral feature modeling according to claim 3 is characterized in that: The step of using the plurality of facets and the radiation flux to add an imaging effect of an optical system to obtain a second grid conversion result includes: Performing fine grid splitting processing on the multiple face elements to obtain a fine grid; Convert the image plane coordinates of the plurality of facets into the refined grid to obtain a first grid conversion result; The imaging effect of the optical system is added to the first grid conversion result in combination with the radiation flux to obtain a second grid conversion result.

5. The detection and identification method based on target spatiotemporal spectral feature modeling according to claim 3 is characterized in that: The performing image plane coordinate conversion processing on the second grid conversion result to obtain the initial target feature data set includes: The second grid conversion result is subjected to fine grid removal through downsampling processing, and the result of the fine grid removal is regressed to the image plane coordinates of the detector to obtain the initial target feature data set.

6. The detection and identification method based on target spatiotemporal spectral feature modeling according to claim 1 is characterized in that: The performing image denoising processing on the to-be-recognized image sequence to obtain the to-be-recognized denoised image sequence comprises: The image sequence to be identified is subjected to image denoising processing by adopting a weighted mean processing method to obtain the denoised image sequence to be identified.

7. The detection and identification method based on target spatiotemporal spectral feature modeling according to claim 1 is characterized in that: The step of generating target candidate screening points through the to-be-identified denoised image sequence comprises: The spatial local contrast and temporal local contrast methods are used to process the noise reduction image sequence to be identified, so as to obtain target candidate points; the target candidate points include the position, pixel distribution and gray value information of each target candidate point; An inversion calculation is performed on the target candidate point to obtain the target candidate screening point.

8. The detection and identification method based on target spatiotemporal spectral feature modeling according to claim 1 is characterized in that: The target identification result is obtained by performing data matching processing using a data matching network based on the target candidate screening points and the target feature database, including: Obtaining a matching range corresponding to the target candidate screening point; Using the data in the target feature database within the matching range as the first target matching database; Performing grayscale quantization and downsampling operations on the first target matching database in sequence to obtain a second target matching database; The data matching network is used to perform data matching processing on the second target matching database and the target candidate screening points to obtain the target recognition result.

9. A detection and identification device based on target spatiotemporal spectral feature modeling, characterized in that: The detection and identification device based on target spatiotemporal spectrum feature modeling includes: an acquisition unit, a preprocessing unit, a screening unit and a data matching unit; The acquisition unit is used to: acquire a sequence of images to be identified; the sequence of images to be identified is a sequence of images captured and transmitted back by the detector; The preprocessing unit is used to: perform image denoising on the image sequence to be identified to obtain a denoised image sequence to be identified; The screening unit is used to: generate target candidate screening points through the noise reduction image sequence to be identified; The data matching unit is used to: perform data matching processing using a data matching network based on the target candidate screening points and the target feature database to obtain a target recognition result; The target feature database is obtained by modeling and simulation using a target surface grid model and multiple sets of simulation parameters; the multiple sets of simulation parameters are obtained by traversing a simulation data set; the simulation data set includes: detector band information and simulated target detection information; the simulated target detection information includes: simulated target surface temperature field distribution, simulated target surface material characteristic parameters, simulated target data affected by external radiation sources, simulated target motion data, detector observation direction and detector observation position.

10. A detection and identification device based on target spatiotemporal spectral feature modeling, characterized in that: include: A processor, a storage medium and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, and when the detection and identification device based on target spatiotemporal spectral feature modeling is running, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to perform the steps of the detection and identification method based on target spatiotemporal spectral feature modeling as described in any one of claims 1-8.

Images