Remote sensing image target recognition method, device, equipment and medium based on multi-channel

Abstract

The present disclosure provides a multi-channel remote sensing image target recognition method, device, equipment and medium. The remote sensing image target recognition method includes: acquiring a remote sensing image, which includes an infrared remote sensing image and a visible light remote sensing image; calibrating the position of the remote sensing image based on the position information of the remote sensing imaging device, and inputting the position into a pre-trained multi-channel recognition network, which includes: extracting the heat distribution characteristics of the infrared remote sensing image and the detail characteristics of the visible light remote sensing image, identifying a preset target based on the heat distribution characteristics and the detail characteristics, and obtaining the relative position coordinates of the preset target in the remote sensing image; and obtaining the latitude and longitude coordinates of the preset target based on the relative position coordinates, the longitude and latitude coordinates of the starting point of the remote sensing image, and the actual distance represented by the distance between adjacent pixels.

Description

Technical Field

[0001] The present disclosure relates to the field of target detection technology, and in particular to a multi-channel remote sensing image target recognition method, device, equipment and medium. Background Art

[0002] With the development of remote sensing technology, intelligent recognition based on remote sensing images has become a trend. However, in remote sensing image recognition, factors such as background interference from objects and cloud cover can lead to high false alarm rates, making it difficult to effectively identify targets. Currently, common remote sensing recognition methods include SAR image recognition, visible light image recognition, and infrared image recognition. From the perspective of imaging mechanisms, SAR, visible light, and infrared images present different image types differently, each with its own advantages and disadvantages. Furthermore, if we want to further subdivide target types for more effective friend-or-foe identification, extracting information using a single imaging method becomes even more challenging, and ensuring recognition accuracy becomes even more difficult.

[0003] Currently, most research institutions still identify remote sensing targets by acquiring optical image information within a specific wavelength band. This approach is relatively convenient for building and maintaining algorithmic code. Code for a specific type of image recognition can be obtained from public code hosting platforms such as GitHub, and then secondary development such as cropping and splicing can be performed to complete the recognition task. However, from the perspective of imaging mechanism, visible light images have a short wavelength and are susceptible to interference from obstacles such as clouds during remote sensing imaging. Furthermore, the large number of channels in the visible light band significantly increases the amount of image information and complicates image recognition. Furthermore, infrared remote sensing images are insensitive to target texture features and color details, making them difficult to classify. Summary of the Invention

[0004] (1) Technical issues to be resolved

[0005] In view of the above problems, the present disclosure provides a multi-channel remote sensing image target recognition method, device, equipment and medium to at least partially solve the current problems of low remote sensing image recognition accuracy, high false alarm rate, poor robustness and susceptibility to interference.

[0006] (2) Technical solution

[0007] On the one hand, the present disclosure provides a multi-channel remote sensing image target recognition method, comprising: acquiring a remote sensing image, wherein the remote sensing image includes an infrared remote sensing image and a visible light remote sensing image; calibrating the position of the remote sensing image based on the position information of the remote sensing imaging device, and inputting the position into a pre-trained multi-channel recognition network, wherein the method comprises: extracting heat distribution features of the infrared remote sensing image and detail features of the visible light remote sensing image, identifying a preset target based on the heat distribution features and detail features, and acquiring the relative position coordinates of the preset target in the remote sensing image; and acquiring the latitude and longitude coordinates of the preset target based on the relative position coordinates, the latitude and longitude coordinates of the starting point of the remote sensing image, and the actual distance represented by the spacing between adjacent pixels.

[0008] According to an embodiment of the present disclosure, the position calibration of the remote sensing image based on the position information of the remote sensing imaging device includes: determining the longitude and latitude coordinates of the starting point of the remote sensing image based on the position information and optical imaging parameters of the remote sensing imaging device; and obtaining the actual distance represented by the spacing between adjacent pixels of the remote sensing image and the longitude and latitude coordinates corresponding to each pixel by solving the longitude and latitude coordinates of the starting point of the remote sensing image and the optical imaging parameters.

[0009] According to an embodiment of the present disclosure, the extracting of heat distribution features of the infrared remote sensing image and detail features of the visible light remote sensing image, and identifying a preset target based on the heat distribution features and detail features includes: extracting features of the slice images corresponding to the infrared remote sensing image and the visible light remote sensing image, and increasing the dimensions to obtain high-dimensional features respectively; based on the high-dimensional features of the infrared remote sensing image and the visible light remote sensing image, using the CONCAT function to perform a merging operation; and identifying the preset target based on the merged high-dimensional features.

[0010] According to an embodiment of the present disclosure, the method further includes: segmenting the infrared remote sensing image and the visible light remote sensing image with a preset resolution and overlap rate to obtain slice images corresponding to the infrared remote sensing image and the visible light remote sensing image.

[0011] According to an embodiment of the present disclosure, the extracting features of the slice images corresponding to the infrared remote sensing image and the visible light remote sensing image, and increasing the dimensions to obtain high-dimensional features respectively include: convolving the slice images through a convolution kernel, iterating, and mapping the slice images into linearly separable high-dimensional features.

[0012] According to an embodiment of the present disclosure, obtaining the relative position coordinates of a preset target in the remote sensing image includes: taking the upper left corner starting point of the remote sensing image as the origin, establishing a two-dimensional plane coordinate system to the right and downward, and quantitatively representing the relative position of the preset target in the remote sensing image based on pixels as the basic unit of measurement.

[0013] According to an embodiment of the present disclosure, it also includes: packaging the infrared remote sensing image, the visible light remote sensing image, the latitude and longitude coordinates of the starting point of the remote sensing image, and the actual distance represented by the distance between adjacent pixels into json format for transmission.

[0014] The second aspect of the present disclosure provides a multi-channel remote sensing image target recognition device, including: an acquisition module, used to acquire remote sensing images within a target area, the remote sensing images including infrared remote sensing images and visible light remote sensing images; a position calibration module, used to calibrate the position of the remote sensing images based on the position information of the remote sensing imaging device, and input the position into a pre-trained multi-channel recognition network; an identification module, used to extract the heat distribution characteristics of the infrared remote sensing image and the detail characteristics of the visible light remote sensing image, identify a preset target based on the heat distribution characteristics and the detail characteristics, and obtain the relative position coordinates of the preset target in the remote sensing image; a position perception module, used to obtain the latitude and longitude coordinates of the preset target based on the relative position coordinates, the latitude and longitude coordinates of the starting point of the remote sensing image, and the actual distance represented by the distance between adjacent pixels.

[0015] The third aspect of the present disclosure provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, each step in the remote sensing image target recognition method is implemented.

[0016] The fourth aspect of the present disclosure further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the computer program implements the various steps in the remote sensing image target recognition method.

[0017] (3) Beneficial effects

[0018] The multi-channel remote sensing image target recognition method, device, equipment and medium provided by the present disclosure integrate infrared and visible light information, pre-process the two and then put them into a multi-channel recognition network, so as to simultaneously judge the heat distribution characteristics and detail characteristics of the predetermined target, thereby greatly improving the recognition accuracy and reducing the recognition difficulty. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] For a more complete understanding of the present disclosure and its advantages, reference will now be made to the following description taken in conjunction with the accompanying drawings, in which:

[0020] Figure 1 The flowchart of the multi-channel remote sensing image target recognition method provided in accordance with an embodiment of the present disclosure is schematically shown;

[0021] Figure 2The following schematically shows a functional architecture diagram of a target detection system provided according to an embodiment of the present disclosure;

[0022] Figure 3 Schematically shows a block diagram of the device and interface connections of the target detection system provided according to an embodiment of the present disclosure;

[0023] Figure 4 The following schematically shows a schematic diagram of the target detection network structure provided according to an embodiment of the present disclosure;

[0024] Figure 5 The schematic diagram of the structure of a dense layer network module involving lightweighting provided in accordance with an embodiment of the present disclosure is shown;

[0025] Figure 6 The following schematically shows a schematic diagram of the residual network structure provided according to an embodiment of the present disclosure;

[0026] Figure 7 The following schematically shows a flow chart of slice push browsing involved in accordance with an embodiment of the present disclosure;

[0027] Figure 8 The following schematically shows a flowchart of a retrieval process for a specified image according to an embodiment of the present disclosure;

[0028] Figure 9 The following schematically shows a structural block diagram of a multi-channel remote sensing image target recognition device provided according to an embodiment of the present disclosure;

[0029] Figure 10 The figure schematically shows a hardware structure diagram of an electronic device. DETAILED DESCRIPTION

[0030] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present disclosure. In the detailed description below, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of well-known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0031] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise," "include," etc. used herein indicate the presence of the features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0032] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0033] The accompanying drawings show some block diagrams and / or flow charts. It should be understood that some blocks in the block diagrams and / or flow charts, or combinations thereof, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when these instructions are executed by the processor, they can create a device for implementing the functions / operations described in the block diagrams and / or flow charts.

[0034] Figure 1 The flowchart of the multi-channel remote sensing image target recognition method provided in accordance with an embodiment of the present disclosure is schematically shown.

[0035] like Figure 1 As shown, the flowchart of the multi-channel remote sensing image target recognition method specifically includes S1-S4.

[0036] In operation S1 , a remote sensing image is acquired, where the remote sensing image includes an infrared remote sensing image and a visible light remote sensing image.

[0037] In operation S2, the remote sensing image is calibrated based on the position information of the remote sensing imaging device and input into a pre-trained multi-channel recognition network.

[0038] In operation S3, heat distribution features of the infrared remote sensing image and detail features of the visible light remote sensing image are extracted, a preset target is identified based on the heat distribution features and detail features, and the relative position coordinates of the preset target in the remote sensing image are obtained.

[0039] In operation S4 , the latitude and longitude coordinates of the preset target are acquired based on the relative position coordinates, the latitude and longitude coordinates of the starting point of the remote sensing image, and the actual distance represented by the distance between adjacent pixels.

[0040] According to an embodiment of the present disclosure, the ground control station is connected to the remote sensing imaging device by wireless communication, and the remote sensing imaging device acquires remote sensing images in real time, and at the same time calibrates the position of the remote sensing images in combination with the device's own position information. The remote sensing imaging device forms a complete remote sensing image file and transmits it to the ground control station. The remote sensing image file includes an infrared remote sensing image, a visible light remote sensing image, the latitude and longitude coordinates of the starting point of the remote sensing image (the first pixel in the upper left corner), and the actual distance represented by the spacing between adjacent pixels. The ground control station parses the complete remote sensing image file, that is, parses and stores the preset target information and the image slices of the preset target, and completes the recognition task of the remote sensing image of the target area through a multi-channel recognition network. Preferably, the infrared detector and the visible light detector in the remote sensing imaging device acquire infrared remote sensing images and visible light remote sensing images respectively.

[0041] According to the embodiments of the present disclosure, compared with a single visible light remote sensing image target recognition method, an additional coaxial infrared remote sensing image is added. The infrared band remote sensing image imaging has better filtering properties for clouds, and some preset targets, such as vehicles, will dissipate a lot of heat during driving. This can effectively obtain the heat distribution characteristics of the target, such as the vehicle's heated tires, engine, exhaust, etc., thereby bringing great advantages to infrared remote sensing image recognition. At the same time, compared with a single infrared remote sensing image target recognition method, the present disclosure utilizes the characteristics of visible light imaging with high granularity, clear texture features, and easy extraction of contour features, which can also effectively reduce the false detection rate of vehicle recognition. The present disclosure utilizes the powerful recognition and classification capabilities of the deep learning model to simultaneously process the above two types of information, taking into account the characteristics of the two types of images during the training process, thereby obtaining more accurate and robust results than those obtained by cross-comparison after single analysis of visible light images, single analysis of infrared images, or single analysis of the above two images.

[0042] Furthermore, operation S2 specifically includes S21-S22.

[0043] In operation S21 , the longitude and latitude coordinates of the starting point of the remote sensing image are determined based on the position information and optical imaging parameters of the remote sensing imaging device.

[0044] In operation S22 , the actual distance represented by the distance between adjacent pixels of the remote sensing image and the longitude and latitude coordinates corresponding to each pixel are obtained by calculating the longitude and latitude coordinates of the starting point of the remote sensing image and the optical imaging parameters.

[0045] According to the embodiments provided by the present disclosure, the optical imaging parameters refer to parameters such as the lens pixel size of the infrared detector and the visible light detector that can be used to calculate the camera target area. By solving the optical imaging parameters, the actual distance represented by the spacing between adjacent pixels is obtained, and combined with the longitude and latitude coordinates of the starting point of the remote sensing image, the longitude and latitude coordinates corresponding to each pixel of the remote sensing image of the target area are obtained, thereby providing assistance for the subsequent acquisition of the actual longitude and latitude coordinates of the preset target. At the same time, the acquired data is compressed and packaged for transmission, which can achieve effective data transmission under narrow bandwidth communication conditions and avoid the problem of insufficient computing power caused by space limitations on the remote sensing imaging device end.

[0046] Furthermore, operation S3 specifically includes S31-S33.

[0047] In operation S31 , features of slice images corresponding to the infrared remote sensing image and the visible light remote sensing image are extracted, and the dimensions are increased to obtain high-dimensional features.

[0048] In operation S32 , a concatenation operation is performed using a CONCAT function based on the high-dimensional features of the infrared remote sensing image and the visible light remote sensing image.

[0049] In operation S33 , a preset target is identified based on the combined high-dimensional features.

[0050] According to an embodiment of the present disclosure, a convolutional neural network is used to extract features from infrared and visible light remote sensing images respectively. After the features of the two types of images are elevated to a higher dimension, the feature information of the two images is combined using a CONCAT operation. Because infrared remote sensing images and visible light remote sensing images are coaxial, when the feature maps of the two are of the same size, the actual physical space corresponding to their high-dimensional feature space is consistent, allowing the model to simultaneously recognize information obtained by the two imaging methods. The two imaging methods are highly complementary at the feature level. Therefore, the present disclosure is more accurate and has stronger anti-interference capabilities than the method of recognizing infrared and visible light remote sensing images separately and then integrating the results.

[0051] Furthermore, operation S3 also includes S34.

[0052] In operation S34 , a two-dimensional plane coordinate system is established with the upper left corner of the remote sensing image as the origin and moving rightward and downward, and the relative position of the preset target in the remote sensing image is quantitatively represented using pixels as the basic measurement unit.

[0053] Commonly used coordinate systems in two-dimensional image measurement include rectangular coordinate systems, polar coordinate systems, and non-orthogonal coordinate systems. Because rectangular coordinate systems can be transformed using linear transformation matrices, they are often used as the basis for coordinate transformation in two-dimensional image measurement. This method establishes a two-dimensional plane coordinate system to determine the relative position of a preset target within a remote sensing image. This plays a crucial role in subsequent measurement and data processing, ensuring the consistency of measurement data and truly achieving measurement of the preset target.

[0054] In the embodiment of the present disclosure, after reading the remote sensing image, the infrared remote sensing image and the visible light remote sensing image are segmented with a preset resolution and overlap ratio to obtain slice images corresponding to the infrared remote sensing image and the visible light remote sensing image.

[0055] Image segmentation is a process of labeling each pixel in an image. This process makes pixels with the same label have certain common visual characteristics. After cutting, each pixel in a sub-region is similar in terms of certain characteristic measurements or calculated characteristics, such as color, brightness, and texture. At the same time, adjacent regions are very different in terms of certain characteristic measurements. The convolutional neural network used in this disclosure removes the fully connected layer at the end, far exceeding traditional methods in accuracy and even efficiency. This disclosure simplifies or changes the image representation by cutting remote sensing images to extract features and measure parameters of preset targets, making higher-level image analysis and understanding possible.

[0056] Furthermore, operation S31 also includes S311.

[0057] In operation S11, the slice image is convolved with a convolution kernel and iteratively mapped into a linearly separable high-dimensional feature.

[0058] In the convolutional neural network's recognition of slice images of remote sensing images, the image is continuously convolved through the convolution kernel, and multiple iterations are performed to map the original image into high-dimensional features. After a certain level of iteration, the features become linearly separable, thereby identifying targets of different categories and further improving the recognition accuracy.

[0059] In the embodiment of the present disclosure, the longitude and latitude coordinates of the starting point of the infrared remote sensing image, the visible light remote sensing image, and the remote sensing image, and the actual distance size represented by the spacing between adjacent pixels are packaged into a json format for transmission. JSON is smaller and faster than XML. Not only is the data format simpler and easier to read and write, the format is compressed, and it occupies less bandwidth, but the interaction is more convenient, easy for people to write and read, and easy for machines to generate and parse. Therefore, in the embodiment of the present disclosure, the longitude and latitude coordinates of the starting point of the infrared remote sensing image, the visible light remote sensing image, and the remote sensing image, and the actual distance size represented by the spacing between adjacent pixels are packaged into a json format for transmission. It has a clear hierarchy and structure, and has a certain degree of versatility, and can provide a reference for the representation of data in other remote sensing image target recognition methods.

[0060] Example 1:

[0061] Remote sensing imaging equipment deploys real-time data acquisition and automatic monitoring background, whose process is named "DataDetSever", and the ground control station deploys data acceptance, decompression and organization management services, whose process is named "DataManSever". The client can also connect to the ground control station through the network, as shown in the attached Figure 2 shown.

[0062] Figure 3 The figure schematically shows a block diagram of the device and interface connection of the target detection system provided according to an embodiment of the present disclosure.

[0063] DataDetSever is deployed on remote sensing imaging equipment and mainly performs the following functions: continuously obtain ground image data from the image payload of the optoelectronic payload device, which is also referred to as the optoelectronic payload. Based on the image input from the optoelectronic payload, it is sent to the position calibration module, and then the remote sensing image file is compressed in tar format and pushed to the ground communication device through the ground communication module. At the same time, it responds to the request for specified data from the DataManServer in an independent thread, and then compresses the specified data output by the intelligent processing terminal and sends it to the intelligent terminal calculation module, and then detects and identifies the target based on the lightweight deep convolutional neural network algorithm, as shown in the attached figure. Figure 4 As shown, the lightweight dense layer network module structure is as follows Figure 5 As shown, the residual module structure is as follows Figure 6 The detection results include the image capture time, image processing time, the relative position coordinates of the preset target in the remote sensing image and its category, and the latitude and longitude coordinates of the preset target. In addition, DataDetSever also responds to control commands issued by DataManServer, such as starting image capture and pausing capture.

[0064] The hardware equipment required for remote sensing imaging equipment mainly includes: a position calibration module, which completes the functions of intelligent calculation, position data acquisition and payload data acquisition; an optoelectronic payload, which completes the functions of ground image data acquisition and transmission and three-axis gimbal stabilization; a front camera, which completes the function of front image acquisition. The image taken by the front camera is sent to the intelligent terminal calculation module, which forwards it to the airborne communication module and realizes the data transmission to the ground control station to assist the remote sensing imaging equipment operator in assisted driving; an airborne communication module, which completes the functions of sending or receiving intelligent calculation-related result data, sending or receiving flight control-related data, and sending or receiving payload-related data; and an external power supply device, namely a power supply module, which mainly supplies power to the airborne equipment, specifically DC 9~15V.

[0065] The functions of the ground control station mainly include three aspects: intelligent terminal computing, data organization and interaction, and route planning. The complete image file compression package push data stream from the DataDetServer is received in the form of an independent background thread, and decompressed to obtain the transmission data of the remote sensing imaging device, and the target detection of the corresponding remote sensing image is completed by relying on the intelligent terminal computing device, and the detection result slice is formed. The detected slice is then geometrically corrected to obtain a clearer image, and all slice information is stored in the database. The original image information is then stored in the database to match the original image retrieval function requirements of the front end, and pushed in the form of websocket. After that, the client connection is responded to, and the slice is displayed on the front end in a dynamic websocket push manner. Respond to the client's retrieval requirements for the specified image, and push the original image retrieval requirements to the DataDetServer of the specified remote sensing imaging device to further obtain the image information of the remote sensing imaging device, as shown in the attached figure. Figure 7 As shown in the following figure, it responds to the needs of data confirmation, management and information export, responds to the needs of map base map display and browsing, and realizes the comprehensive arrangement of information, as shown in the following figure. Figure 8 shown.

[0066] The hardware equipment of the ground control station mainly includes: data organization and interaction server: realizing data organization, interface interaction, online binding of flight mission planning waypoints and payload control functions; ground communication equipment: realizing the functions of sending / receiving intelligent calculation related result data, sending / receiving flight control related data and sending / receiving payload related data; flight remote control equipment: realizing flight operation functions; external power supply equipment: providing power for ground equipment.

[0067] According to the method provided in the embodiments of the present disclosure, when identifying targets in remote sensing images, a multi-channel recognition network is used to simultaneously identify infrared remote sensing images and visible light remote sensing images, so that the images transmitted back in the system are no longer images of a single category, and the target recognition network structure is also optimized accordingly, thereby having higher detection and recognition capabilities, and solving the problems of low target detection accuracy and high false detection rate in the existing technology.

[0068] Based on the above remote sensing image target recognition method, the present disclosure also provides a remote sensing image target recognition device based on multiple channels. Figure 9 The device is described in detail.

[0069] like Figure 9 As shown, the multi-channel remote sensing image target recognition device of this embodiment includes an acquisition module 901 , a position calibration module 902 , a recognition module 903 and a position perception module 904 .

[0070] The acquisition module 901 is used to acquire remote sensing images within the target area. The remote sensing images include infrared remote sensing images and visible light remote sensing images.

[0071] The position calibration module 902 is used to calibrate the position of the remote sensing image based on the position information of the remote sensing imaging device and input it into the pre-trained multi-channel recognition network.

[0072] The recognition module 903 is used to extract the heat distribution features of the infrared remote sensing image and the detail features of the visible light remote sensing image, identify the preset target based on the heat distribution features and detail features, and obtain the relative position coordinates of the preset target in the remote sensing image.

[0073] The position sensing module 904 is used to obtain the latitude and longitude coordinates of a preset target based on the relative position coordinates, the latitude and longitude coordinates of the starting point of the remote sensing image, and the actual distance represented by the distance between adjacent pixels.

[0074] According to an embodiment of the present disclosure, any multiple modules among the acquisition module 901, the position calibration module 902, the identification module 903 and the position perception module 904 can be combined into one module for implementation, or any one of the modules can be split into multiple modules. Alternatively, at least part of the functions of one or more of these modules can be combined with at least part of the functions of other modules and implemented in one module. According to an embodiment of the present disclosure, at least one of the acquisition module 901, the position calibration module 902, the identification module 903 and the position perception module 904 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application specific integrated circuit (ASIC), or can be implemented by hardware or firmware such as any other reasonable way of integrating or packaging the circuit, or implemented in any one of the three implementation methods of software, hardware and firmware or in an appropriate combination of any of them. Alternatively, at least one of the acquisition module 901, the position calibration module 902, the identification module 903 and the position perception module 904 can be at least partially implemented as a computer program module, and when the computer program module is run, the corresponding function can be executed.

[0075] See Figure 10 , Figure 10 A hardware structure diagram of an electronic device is shown.

[0076] The electronic device described in this embodiment includes:

[0077] The memory 101, the processor 102 and the computer program stored in the memory 101 and capable of running on the processor, when the processor executes the program, realize the aforementioned Figure 1 The multi-channel remote sensing image target recognition method described in the illustrated embodiment.

[0078] Furthermore, the electronic device further includes:

[0079] At least one input device 103 ; at least one output device 104 .

[0080] The memory 101 , processor 102 , input device 103 and output device 104 are connected via a bus 105 .

[0081] The input device 103 may be a camera, a touch panel, a physical button, a mouse, etc. The output device 104 may be a display screen.

[0082] The memory 101 can be a high-speed random access memory (RAM) memory or a non-volatile memory such as a disk drive. The memory 101 is used to store a set of executable program codes. The processor 102 is coupled to the memory 101.

[0083] Furthermore, an embodiment of the present invention further provides a computer-readable storage medium, which may be provided in the terminal in each of the above embodiments. The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the above Figure 1 The multi-channel remote sensing image target recognition method described in the illustrated embodiment. Furthermore, the computer storage medium may be any medium capable of storing program code, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0084] In the multiple embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the embodiments described above are merely illustrative. For example, the division of the modules is merely a logical function division. In actual implementation, there may be other division methods, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication link shown or discussed can be through some interfaces, and the indirect coupling or communication link of the modules can be electrical, mechanical or other forms.

[0085] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules may be selected to achieve the purpose of the present embodiment according to actual needs.

[0086] In addition, the functional modules in various embodiments of the present invention may be integrated into a single processing module, or each module may exist physically separately, or two or more modules may be integrated into a single module. The aforementioned integrated modules may be implemented in the form of hardware or software functional modules.

[0087] It should be noted that for the aforementioned method embodiments, for ease of description, they are all expressed as a series of action combinations. However, those skilled in the art should be aware that the present invention is not limited by the order of the actions described, because according to the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0088] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0089] The above is a description of the method, device, electronic device, and storage medium for constructing a user psychological model provided by the present invention. For those skilled in the art, according to the ideas of the embodiments of the present invention, there may be changes in the specific implementation methods and application scopes. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A multi-channel remote sensing image target recognition method, characterized in that: include: Acquiring remote sensing images, wherein the remote sensing images include infrared remote sensing images and visible light remote sensing images; The remote sensing image is calibrated based on the position information of the remote sensing imaging device and input into a pre-trained multi-channel recognition network, which includes: Extracting heat distribution features of the infrared remote sensing image and detail features of the visible light remote sensing image, and identifying a preset target based on the heat distribution features and detail features; specifically comprising: segmenting the infrared remote sensing image and the visible light remote sensing image at a preset resolution and overlap rate to obtain slice images corresponding to the infrared remote sensing image and the visible light remote sensing image; convolving the slice images with a convolution kernel, iteratively mapping the slice images into linearly separable high-dimensional features; performing a merging operation using a CONCAT function based on the high-dimensional features of the infrared remote sensing image and the visible light remote sensing image; and identifying the preset target based on the merged high-dimensional features; The relative position coordinates of the preset target in the remote sensing image are obtained, and the latitude and longitude coordinates of the preset target are obtained based on the relative position coordinates, the latitude and longitude coordinates of the starting point of the remote sensing image, and the actual distance represented by the adjacent pixel spacing.

2. The remote sensing image target recognition method according to claim 1, characterized in that: The performing position calibration on the remote sensing image based on the position information of the remote sensing imaging device includes: Determining the latitude and longitude coordinates of the starting point of the remote sensing image based on the position information and optical imaging parameters of the remote sensing imaging device; By calculating the latitude and longitude coordinates of the starting point of the remote sensing image and the optical imaging parameters, the actual distance represented by the distance between adjacent pixels of the remote sensing image and the latitude and longitude coordinates corresponding to each pixel are obtained.

3. The remote sensing image target recognition method according to claim 1, characterized in that: The acquiring of relative position coordinates of a preset target in the remote sensing image comprises: A two-dimensional plane coordinate system is established with the upper left corner starting point of the remote sensing image as the origin, and pixels are used as the basic measurement unit to quantitatively represent the relative position of the preset target in the remote sensing image.

4. The remote sensing image target recognition method according to claim 1, characterized in that: Also includes: The infrared remote sensing image, the visible light remote sensing image, the latitude and longitude coordinates of the starting point of the remote sensing image, and the actual distance represented by the distance between adjacent pixels are packaged into json format for transmission.

5. A multi-channel remote sensing image target recognition device, characterized in that: include: An acquisition module is used to acquire remote sensing images within the target area, wherein the remote sensing images include infrared remote sensing images and visible light remote sensing images; A position calibration module, configured to calibrate the position of the remote sensing image based on the position information of the remote sensing imaging device and input the calibration information into a pre-trained multi-channel recognition network; The recognition module is configured to extract heat distribution features of the infrared remote sensing image and detail features of the visible light remote sensing image, and identify a preset target based on the heat distribution features and detail features. The recognition module specifically comprises: slicing the infrared remote sensing image and the visible light remote sensing image at a preset resolution and overlap ratio to obtain slice images corresponding to the infrared remote sensing image and the visible light remote sensing image; iteratively convolving the slice images with a convolution kernel to map the slice images into linearly separable high-dimensional features; performing a merging operation using a CONCAT function based on the high-dimensional features of the infrared remote sensing image and the visible light remote sensing image; and identifying the preset target based on the merged high-dimensional features. The position perception module is used to obtain the relative position coordinates of the preset target in the remote sensing image, and obtain the latitude and longitude coordinates of the preset target based on the relative position coordinates, the latitude and longitude coordinates of the starting point of the remote sensing image and the actual distance represented by the distance between adjacent pixels.

6. An electronic device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, each step of the remote sensing image target recognition method according to any one of claims 1 to 4 is implemented.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, each step of the remote sensing image target recognition method according to any one of claims 1 to 4 is implemented.

Images