A remote sensing image target recognition method and system based on artificial intelligence

Abstract

The present application relates to a method and system for remote sensing image target recognition based on artificial intelligence, the method comprising: in response to the acquired remote sensing image, preprocessing the remote sensing image to obtain a reference remote sensing image, the preprocessing comprising noise reduction, image correction and image cropping; performing range extraction in the reference remote sensing image to obtain a first region of interest; performing object extraction in the first region of interest to obtain at least one feature object; judging the integrity of the feature object and determining the missing position of the feature object; creating a second region of interest based on the missing position and searching in the second region of interest to find a covered object, the type of the covered object being the same as the type of the covered feature object. The method and system for remote sensing image target recognition based on artificial intelligence disclosed in the present application uses range extraction and supplementary search and extraction means for missing positions to realize the recognition of overlapping targets, so as to improve the number recognition accuracy of similar targets.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a remote sensing image target recognition method and system based on artificial intelligence. Background Art

[0002] Remote sensing image target recognition is mainly based on the theory and methods of pattern recognition, that is, using computers or other devices to automatically identify objects, graphics and other information in images. In the field of remote sensing, this usually involves the extraction and analysis of the spectral characteristics, texture characteristics, spatial characteristics, and temporal characteristics of objects in remote sensing images, so as to achieve the recognition of target objects.

[0003] At present, small unmanned terminals in the coverage area have begun to carry remote sensing equipment for active reconnaissance to improve the perception of the coverage area. The advantages of using remote sensing for reconnaissance are data richness and long-distance reconnaissance, but there are certain difficulties in data processing. For example, the current problems of unrecognition and quantity recognition errors caused by the overlap of similar targets require further research. Summary of the invention

[0004] The present application provides a remote sensing image target recognition method and system based on artificial intelligence, which uses range extraction and supplementary retrieval and extraction means for missing positions to realize the recognition of overlapping targets, so as to improve the accuracy of quantitative recognition of similar targets.

[0005] The above-mentioned purpose of the present application is achieved through the following technical solutions:

[0006] In a first aspect, the present application provides a remote sensing image target recognition method based on artificial intelligence, comprising:

[0007] In response to the acquired remote sensing image, preprocessing the remote sensing image to obtain a reference remote sensing image, wherein the preprocessing includes noise reduction, image correction and image cropping;

[0008] Performing range extraction in the reference remote sensing image to obtain a first region of interest, wherein the number of the first region of interest is at least one;

[0009] Extracting objects in the first region of interest to obtain at least one feature object;

[0010] Determine the completeness of feature objects and determine the missing locations of feature objects;

[0011] A second region of interest is created based on the missing position and a search is performed within the second region of interest to find a masked object, where the type of the masked object is the same as the type of the masked feature object.

[0012] In a possible implementation manner of the first aspect, performing range extraction in a reference remote sensing image and obtaining a first region of interest includes:

[0013] Using edge detection to extract the range in the remote sensing image, a suspected first area of ​​interest is obtained;

[0014] Divide the content included in the suspected first area of ​​interest into blocks and determine the type of each block;

[0015] The suspected first region of interest is reduced using the block type to obtain the first region of interest.

[0016] In a possible implementation manner of the first aspect, reducing the suspected first region of interest by using the type of block includes:

[0017] Determine at least one collection area according to the collection degree of the block type and integrate the collection areas of the same type, which is recorded as the first integrated collection area;

[0018] Incorporating other blocks located inside the first integrated collection area into the first integrated collection area to obtain a second integrated collection area;

[0019] The second integrated pooling area is used as the first area of ​​interest;

[0020] When integrating the collection areas of the same type, it is required that the enclosed area of ​​the collection areas of the same type is the smallest.

[0021] In a possible implementation manner of the first aspect, creating the second region of interest based on the missing position includes:

[0022] Randomly select multiple feature points on the edge of the feature object;

[0023] Calculate the eigenvalues ​​of the feature points, which include the color mean and the degree of color change in a certain direction;

[0024] An attempt is made to reconstruct the missing position of the feature object according to the feature values ​​of the feature points and a second region of interest is created at the reconstructed missing position.

[0025] In a possible implementation manner of the first aspect, searching in the second area of ​​interest includes:

[0026] Create a selection region at each of the two breakpoints of the missing position and generate a selection operator in the selection region;

[0027] Use the selection operator to search in the second region of interest to obtain a reference demarcation point;

[0028] The obtained reference dividing point is used to generate a dividing line, and the two end points of the dividing line coincide with the two breakpoints of the missing position respectively;

[0029] Among them, all the contents included in the selection area participate in generating the selection operator.

[0030] In a possible implementation of the first aspect, when generating a selection operator, the generated selection operator includes two groups of pixel points, the difference between the same group of pixel points is within a first allowable range, and the difference on both sides of the boundary line between the two groups of pixel points is within a second allowable range.

[0031] In a possible implementation manner of the first aspect, when the selection operator cannot be generated, the method further includes:

[0032] The selected area is processed in layers to obtain multiple sub-selected areas, each of which corresponds to a color interval;

[0033] A selection operator is generated using the content included in the sub-selection area, where the selection operator includes pixel points and blank points.

[0034] In a second aspect, the present application provides a remote sensing image target recognition device based on artificial intelligence, comprising:

[0035] A preprocessing unit, for preprocessing the remote sensing image in response to the acquired remote sensing image to obtain a reference remote sensing image, wherein the preprocessing includes noise reduction, image correction and image cropping;

[0036] A range extraction unit is used to perform range extraction in the reference remote sensing image to obtain a first region of interest, where the number of the first region of interest is at least one;

[0037] An object extraction unit, configured to extract objects in the first region of interest to obtain at least one feature object;

[0038] A judging unit, used to judge the completeness of the feature object and determine the missing position of the feature object;

[0039] The retrieval unit is used to create a second region of interest based on the missing position and perform a search in the second region of interest to find a covered object, where the type of the covered object is the same as the type of the covered feature object.

[0040] In a third aspect, the present application provides a remote sensing image target recognition system based on artificial intelligence, the system comprising:

[0041] one or more memories for storing instructions; and

[0042] One or more processors, used to call and run the instructions from the memory to execute the method as described in the first aspect and any possible implementation of the first aspect.

[0043] In a fourth aspect, the present application provides a computer-readable storage medium, the computer-readable storage medium comprising:

[0044] Program, when the program is executed by a processor, the method described in the first aspect and any possible implementation of the first aspect is executed.

[0045] In a fifth aspect, the present application provides a computer program product, comprising program instructions. When the program instructions are executed by a computing device, the method described in the first aspect and any possible implementation of the first aspect is executed.

[0046] In a sixth aspect, the present application provides a chip system, which includes a processor for implementing the functions involved in the above aspects, for example, generating, receiving, sending, or processing the data and / or information involved in the above methods.

[0047] The chip system may be composed of chips, or may include chips and other discrete devices.

[0048] In a possible design, the chip system also includes a memory, which is used to store necessary program instructions and data. The processor and the memory can be decoupled and respectively set on different devices, connected by wired or wireless means, or the processor and the memory can also be coupled on the same device. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 This is a schematic diagram of the steps of a remote sensing image target recognition method based on artificial intelligence provided in this application.

[0050] Figure 2 This is a schematic diagram provided in the present application for dividing a reference remote sensing image into regions using color or texture features.

[0051] Figure 3 This is a schematic diagram provided in the present application for dividing a reference remote sensing image into regions using an edge detection method.

[0052] Figure 4 It is a schematic diagram of a missing location of a feature object provided by the present application.

[0053] Figure 5 It is a schematic diagram of generating a dividing line provided by the present application. DETAILED DESCRIPTION

[0054] The technical solution in this application is further described in detail below in conjunction with the accompanying drawings.

[0055] This application discloses a remote sensing image target recognition method based on artificial intelligence, please refer to Figure 1 In some examples, the artificial intelligence-based remote sensing image target recognition method disclosed in this application includes the following steps:

[0056] S101, in response to the acquired remote sensing image, preprocessing the remote sensing image to obtain a reference remote sensing image, the preprocessing including noise reduction, image correction and image cropping;

[0057] S102, performing range extraction in the reference remote sensing image to obtain a first region of interest, where the number of the first region of interest is at least one;

[0058] S103, extracting objects within the first area of ​​interest to obtain at least one feature object;

[0059] S104, judging the completeness of the feature object and determining the missing position of the feature object;

[0060] S105, creating a second region of interest based on the missing position and searching in the second region of interest to find a covering object, where the type of the covering object is the same as the type of the covered feature object.

[0061] In step S101, the acquired remote sensing image is first preprocessed. The preprocessing includes three parts: noise reduction, image correction and image cropping. The specific contents are as follows:

[0062] Noise reduction of remote sensing images is an important part of remote sensing image processing. Due to various factors such as sensor problems, atmospheric interference, surface reflection, etc., various types of noise often exist in remote sensing images, which will affect the quality of the image and subsequent application effects. Therefore, it is very necessary to perform noise reduction on remote sensing images.

[0063] Noise type: Noise in remote sensing images mainly includes periodic noise, spike noise, stripe noise (bad lines), etc. Periodic noise is usually caused by mechanical vibration of the sensor or scanning system, and appears as a series of repeated interference patterns. Spike noise is caused by abnormal values ​​or mutations of a single pixel, and appears as isolated bright or dark spots. Stripe noise is stripe-shaped noise parallel to the scanning direction, which is usually related to sensor performance or radiation correction.

[0064] Noise reduction processing method:

[0065] Bandpass filtering or slot filtering:

[0066] Principle: Use filters to pass signals within a specific frequency range and suppress noise in other frequency ranges.

[0067] Application: Suitable for eliminating periodic noise. By adjusting the filter parameters, the interference pattern corresponding to the noise frequency can be selectively eliminated.

[0068] Fourier Transform:

[0069] Principle: Convert the image from the spatial domain to the frequency domain, use the characteristics of the frequency domain to perform filtering, and then convert it back to the spatial domain.

[0070] Application: Suitable for eliminating spike noise and strip noise. Through Fourier transform, the noise can be separated in the frequency domain and suppressed by low-pass filter or band-stop filter.

[0071] Mean filter:

[0072] Principle: Take the average value of pixels within a certain range around the pixel to be processed as the new value of the pixel.

[0073] Application: Suitable for smoothing images and reducing the impact of random noise. However, it should be noted that mean filtering may cause loss of image details.

[0074] Median filter:

[0075] Principle: Take the median value of the pixels within a certain range around the pixel to be processed as the new value of the pixel.

[0076] Application: Suitable for removing salt and pepper noise (i.e. randomly distributed bright and dark spots). Median filtering is better than mean filtering in preserving image edge information.

[0077] Edge-preserving smoothing filter:

[0078] Principle: Keep the edge information of the image while smoothing it. This is usually achieved by judging whether the current pixel is an edge point. If it is an edge point, no smoothing is performed.

[0079] Application: Suitable for scenes where image details need to be preserved.

[0080] Mathematical morphology filtering:

[0081] Principle: Use mathematical morphology's opening and closing operations to eliminate noise in images.

[0082] Application: Suitable for eliminating small bright spots or dark spots in images. Mathematical morphological filtering has a good effect in maintaining image structural features.

[0083] Evaluation of noise reduction processing effect: The effect of noise reduction processing is usually evaluated by comparing the image quality before and after processing. Common evaluation indicators include signal-to-noise ratio (SNR), mean square error (MSE), peak signal-to-noise ratio (PSNR), etc.

[0084] The remote sensing image is obtained by obtaining a reference remote sensing image, and then in step S102, a range is extracted in the reference remote sensing image to obtain a first region of interest. The number of the first region of interest may be one or more, and the first region of interest refers to a region where a feature object may exist.

[0085] The specific method of range extraction is to use color or texture feature segmentation or edge detection.

[0086] See also Figure 2 The specific process of using color or texture feature segmentation is to first divide the reference remote sensing image into multiple squares, and then calculate the color features or texture features in each square. The color features or texture features are obtained based on the sampling of the coverage area. When the color features or texture features of a square do not meet the requirements, it will be included in the first area of ​​interest.

[0087] See also Figure 3 The specific process of using edge detection is to directly extract edges in the reference remote sensing image. At this time, multiple contours will be obtained, and then the contour shapes will be judged, and the areas corresponding to the suspected contours will be included in the first area of ​​interest.

[0088] Then, objects are extracted in the first region of interest to obtain at least one feature object, that is, the content in step S103. The specific method of extracting objects and obtaining feature objects is to divide the first region of interest into multiple squares, and then send each square to the comparison model for comparison or use the random forest method for voting to obtain the content included in each square.

[0089] Then the results are summarized, and multiple results such as personnel, weapons, vehicles, etc. may be obtained.

[0090] Because the grid-dividing processing method is used in the above judgment process, the integrity of the feature object may be incomplete, which is the problem of similar target overlap mentioned in the background technology of this application. The overlap of similar targets does not affect the summary results when the results are summarized, but the problem of incorrect quantitative results will occur.

[0091] Therefore, in step S104, the integrity of the feature object is determined and the missing position of the feature object is determined. Figure 4 As shown), step S105 is executed, in which a second region of interest is created based on the missing position and a search is performed in the second region of interest to find a covering object, the type of which is the same as the type of the covered feature object.

[0092] Performing a secondary search for masked objects can solve the problem of incorrect quantity results.

[0093] In some examples, a specific method of performing range extraction in a reference remote sensing image and obtaining a first area of ​​interest is:

[0094] S201, performing range extraction in a reference remote sensing image using an edge detection method to obtain a suspected first region of interest;

[0095] S202, dividing the content included in the suspected first interest region into blocks and determining the type of each block;

[0096] S203: Reduce the suspected first region of interest by using a block type to obtain a first region of interest.

[0097] In step S201 to step S203, the edge detection method is first used to perform range extraction in the benchmark remote sensing image. This step corresponds to the processing method of dividing the benchmark remote sensing image into regions mentioned in the aforementioned content. Because in this method, some squares have interference content, the purpose of the edge detection method is to remove these interference contents.

[0098] Then, the type is determined by using a block processing method. The specific process is to integrate blocks of the same type and discard blocks of undetermined type. This step can reduce the first region of interest to obtain the first region of interest.

[0099] In some examples, the specific method of using the block type to reduce the suspected first region of interest is as follows:

[0100] Determine at least one collection area according to the collection degree of the block type and integrate the collection areas of the same type, which is recorded as the first integrated collection area;

[0101] Incorporating other blocks located inside the first integrated collection area into the first integrated collection area to obtain a second integrated collection area;

[0102] The second integrated pooling area is used as the first area of ​​interest;

[0103] When integrating the collection areas of the same type, it is required that the enclosed area of ​​the collection areas of the same type is the smallest.

[0104] This part corresponds to the situation where the block distribution is relatively scattered. For this situation, the integration processing method is used. The specific process is to first integrate the collection areas of the same type, and then incorporate the other blocks located inside the first integration collection area into the first integration collection area to obtain the second integration collection area. At this time, the enclosed area of ​​the collection areas of the same type needs to be minimized, in order to introduce as few interference factors as possible.

[0105] In some examples, the second region of interest is created based on the missing location by:

[0106] S301, randomly selecting multiple feature points on the edge of the feature object;

[0107] S302, calculating the characteristic value of the characteristic point, the characteristic value includes the color mean and the color change degree in a certain direction;

[0108] S303: Attempt to reconstruct the missing position of the feature object according to the feature value of the feature point and create a second region of interest at the reconstructed missing position.

[0109] In step S301 to step S303, a plurality of feature points are first randomly selected on the edge of the feature object, and then the feature values ​​of the feature points are calculated. The feature values ​​have two dimensions, namely, the color mean and the color variation degree in a certain direction.

[0110] The color mean refers to the average value of the colors corresponding to multiple feature points, and grayscale value is generally used here instead. The color change degree in a certain direction refers to the fact that the color value of the feature point may change in a search direction.

[0111] After determining the search direction, if the color values ​​of multiple randomly selected feature points are stable, then the search is performed based on this. If there is a color change, then the search is performed based on the color change. When searching based on the color change, it is necessary to calculate the color increase rate or color decrease rate based on the distance.

[0112] In some examples, the specific method of searching in the second area of ​​interest is:

[0113] S401, creating a selection region at two breakpoints of the missing position respectively and generating a selection operator in the selection region;

[0114] S402, using a selection operator to search in the second region of interest to obtain a reference demarcation point;

[0115] S403, using the obtained reference demarcation point to generate a demarcation line, wherein two endpoints of the demarcation line coincide with two breakpoints of the missing position respectively;

[0116] Among them, all the contents included in the selection area participate in generating the selection operator.

[0117] In the above steps, firstly, a selection area is created at the two breakpoints of the missing position respectively, and then a selection operator is generated using the content included in the selection area. The selection operator is generally a rectangle with a side length of MxN, where M and N are both natural numbers greater than zero.

[0118] In some possible implementations, the selection area is generally circular or rectangular.

[0119] The specific method of using the selection operator to search in the second area of ​​interest is to drive the selection operator to move in the second area of ​​interest, and then operate the selection operator with the content moved to the corresponding position. The specific method of the operation is to calculate the similarity of the two matrices. If the similarity of the two matrices meets the requirements, then it is considered that the reference dividing point is obtained at this time.

[0120] The pixel points included in the rectangle corresponding to the selection operator, whose positions and values ​​constitute a matrix.

[0121] See also Figure 5 Finally, the reference demarcation point is used to generate the demarcation line. At this time, the two endpoints of the demarcation line need to coincide with the two breakpoints of the missing position. The reason for adding requirements is that the generated demarcation line may be wrong, and these wrong demarcation lines need to be excluded using requirements.

[0122] In some possible implementations, when generating a selection operator, the generated selection operator includes two groups of pixel points, the difference between the same group of pixel points is within a first allowed range, and the difference on both sides of the boundary line between the two groups of pixel points is within a second allowed range.

[0123] When the selection operator cannot be generated, use the following method to handle it:

[0124] The selected area is processed in layers to obtain multiple sub-selected areas, each of which corresponds to a color interval;

[0125] A selection operator is generated using the content included in the sub-selection area, where the selection operator includes pixel points and blank points.

[0126] The present application also provides a remote sensing image target recognition device based on artificial intelligence, comprising:

[0127] A preprocessing unit, for preprocessing the remote sensing image in response to the acquired remote sensing image to obtain a reference remote sensing image, wherein the preprocessing includes noise reduction, image correction and image cropping;

[0128] A range extraction unit is used to perform range extraction in the reference remote sensing image to obtain a first region of interest, where the number of the first region of interest is at least one;

[0129] An object extraction unit, configured to extract objects in the first region of interest to obtain at least one feature object;

[0130] A judging unit, used to judge the completeness of the feature object and determine the missing position of the feature object;

[0131] The retrieval unit is used to create a second region of interest based on the missing position and perform a search in the second region of interest to find a covered object, where the type of the covered object is the same as the type of the covered feature object.

[0132] Further, performing range extraction in the reference remote sensing image and obtaining the first region of interest includes:

[0133] Using edge detection to extract the range in the remote sensing image, a suspected first area of ​​interest is obtained;

[0134] Divide the content included in the suspected first area of ​​interest into blocks and determine the type of each block;

[0135] The suspected first region of interest is reduced using the block type to obtain the first region of interest.

[0136] Further, using the type of block to reduce the suspected first region of interest includes:

[0137] Determine at least one collection area according to the collection degree of the block type and integrate the collection areas of the same type, which is recorded as the first integrated collection area;

[0138] Incorporating other blocks located inside the first integrated collection area into the first integrated collection area to obtain a second integrated collection area;

[0139] The second integrated pooling area is used as the first area of ​​interest;

[0140] When integrating the collection areas of the same type, it is required that the enclosed area of ​​the collection areas of the same type is the smallest.

[0141] Further, creating a second region of interest based on the missing position includes:

[0142] Randomly select multiple feature points on the edge of the feature object;

[0143] Calculate the eigenvalues ​​of the feature points, which include the color mean and the degree of color change in a certain direction;

[0144] An attempt is made to reconstruct the missing position of the feature object according to the feature values ​​of the feature points and a second region of interest is created at the reconstructed missing position.

[0145] Further, searching in the second area of ​​interest includes:

[0146] Create a selection region at each of the two breakpoints of the missing position and generate a selection operator in the selection region;

[0147] Use the selection operator to search in the second region of interest to obtain a reference demarcation point;

[0148] The obtained reference dividing point is used to generate a dividing line, and the two end points of the dividing line coincide with the two breakpoints of the missing position respectively;

[0149] Among them, all the contents included in the selection area participate in generating the selection operator.

[0150] Further, when generating the selection operator, the generated selection operator includes two groups of pixel points, the difference between the pixel points in the same group is within a first allowable range, and the difference between the two sides of the boundary line of the two groups of pixel points is within a second allowable range.

[0151] Furthermore, when the selection operator cannot be generated, it also includes:

[0152] The selected area is processed in layers to obtain multiple sub-selected areas, each of which corresponds to a color interval;

[0153] A selection operator is generated using the content included in the sub-selection area, where the selection operator includes pixel points and blank points.

[0154] In one example, the unit in any of the above devices can be one or more integrated circuits configured to implement the above methods, such as: one or more application specific integrated circuits (ASICs), or, one or more digital signal processors (DSPs), or, one or more field programmable gate arrays (FPGAs), or a combination of at least two of these integrated circuit forms.

[0155] For another example, when the units in the device can be implemented in the form of a processing element scheduling program, the processing element can be a general-purpose processor, such as a central processing unit (CPU) or other processor that can call a program. For another example, these units can be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0156] Various objects such as various messages / information / equipment / network elements / systems / devices / actions / operations / processes / concepts that may appear in this application are named. It can be understood that these specific names do not constitute a limitation on the relevant objects. The names assigned may change with factors such as scenarios, contexts or usage habits. The understanding of the technical meaning of the technical terms in this application should be mainly determined from the functions and technical effects embodied / executed in the technical scheme.

[0157] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0158] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units 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 connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0159] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0160] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0161] It should also be understood that in various embodiments of the present application, the first, second, etc. are only used to indicate that multiple objects are different. For example, the first time window and the second time window are only used to indicate different time windows. They should not have any impact on the time window itself, and the first, second, etc. mentioned above should not impose any limitations on the embodiments of the present application.

[0162] It should also be understood that in the various embodiments of the present application, unless otherwise specified or there is a logical conflict, the terms and / or descriptions between different embodiments are consistent and can be referenced to each other, and the technical features in different embodiments can be combined to form new embodiments according to their internal logical relationships.

[0163] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a computer-readable storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned computer-readable storage medium includes: various media that can store program codes, 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.

[0164] The present application also provides a remote sensing image target recognition system based on artificial intelligence, the system comprising:

[0165] one or more memories for storing instructions; and

[0166] One or more processors are used to call and run the instructions from the memory to execute the method as described above.

[0167] The present application also provides a computer program product, which includes instructions. When the instructions are executed, the terminal device and the network device perform operations of the terminal device and the network device corresponding to the above method.

[0168] The present application also provides a chip system, which includes a processor for implementing the functions involved in the above content, such as generating, receiving, sending, or processing the data and / or information involved in the above method.

[0169] The chip system may be composed of chips, or may include chips and other discrete devices.

[0170] The processor mentioned in any of the above places can be a CPU, a microprocessor, an ASIC, or one or more integrated circuits for executing programs for controlling the above-mentioned feedback information transmission method.

[0171] In a possible design, the chip system also includes a memory, which is used to store necessary program instructions and data. The processor and the memory can be decoupled and respectively set on different devices, connected by wire or wireless means to support the chip system to implement various functions in the above embodiments. Alternatively, the processor and the memory can also be coupled on the same device.

[0172] Optionally, the computer instructions are stored in a memory.

[0173] Optionally, the memory is a storage unit within the chip, such as a register, a cache, etc. The memory can also be a storage unit within the terminal located outside the chip, such as a ROM or other types of static storage devices that can store static information and instructions, RAM, etc.

[0174] It can be understood that the memory in the present application can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories.

[0175] The non-volatile memory may be a ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), or a flash memory.

[0176] The volatile memory may be a RAM, which is used as an external cache. There are many different types of RAM, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), and direct memory bus RAM.

[0177] The embodiments of this specific implementation method are all preferred embodiments of the present application, and are not intended to limit the protection scope of the present application. Therefore, all equivalent changes made based on the structure, shape, and principle of the present application should be included in the protection scope of the present application.

Claims

1. A remote sensing image target recognition method based on artificial intelligence, characterized in that: include: In response to the acquired remote sensing image, preprocessing the remote sensing image to obtain a reference remote sensing image, wherein the preprocessing includes noise reduction, image correction and image cropping; Performing range extraction in the reference remote sensing image to obtain a first region of interest, wherein the number of the first region of interest is at least one; Extracting objects in the first region of interest to obtain at least one feature object; Determine the completeness of feature objects and determine the missing locations of feature objects; Creating a second region of interest based on the missing position and searching within the second region of interest to find a masked object, where the type of the masked object is the same as the type of the masked feature object; Creating a second region of interest based on the missing location includes: Randomly select multiple feature points on the edge of the feature object; Calculate the eigenvalues ​​of the feature points, which include the color mean and the degree of color change in a certain direction; An attempt is made to reconstruct the missing position of the feature object according to the feature values ​​of the feature points and a second region of interest is created at the reconstructed missing position.

2. The method for remote sensing image target recognition based on artificial intelligence according to claim 1, characterized in that: Extracting the range in the reference remote sensing image and obtaining the first area of ​​interest includes: Using edge detection to extract the range in the remote sensing image, a suspected first area of ​​interest is obtained; Divide the content included in the suspected first area of ​​interest into blocks and determine the type of each block; The suspected first region of interest is reduced using the block type to obtain the first region of interest.

3. The method for remote sensing image target recognition based on artificial intelligence according to claim 2 is characterized in that: The types of blocks used to reduce the suspected first area of ​​interest include: Determine at least one collection area according to the collection degree of the block type and integrate the collection areas of the same type, which is recorded as the first integrated collection area; Incorporating other blocks located inside the first integrated collection area into the first integrated collection area to obtain a second integrated collection area; The second integrated pooling area is used as the first area of ​​interest; When integrating the collection areas of the same type, it is required that the enclosed area of ​​the collection areas of the same type is the smallest.

4. The method for remote sensing image target recognition based on artificial intelligence according to claim 1, characterized in that: Searching within the second area of ​​interest includes: Create a selection region at each of the two breakpoints of the missing position and generate a selection operator in the selection region; Use the selection operator to search in the second region of interest to obtain a reference demarcation point; The obtained reference dividing point is used to generate a dividing line, and the two end points of the dividing line coincide with the two breakpoints of the missing position respectively; Among them, all the contents included in the selection area participate in generating the selection operator.

5. The method for remote sensing image target recognition based on artificial intelligence according to claim 1, characterized in that: When generating a selection operator, the generated selection operator includes two groups of pixel points, the difference between the same group of pixel points is within a first allowable range, and the difference on both sides of the boundary line between the two groups of pixel points is within a second allowable range.

6. The method for remote sensing image target recognition based on artificial intelligence according to claim 5, characterized in that: When the selection operator cannot be generated, it also includes: The selected area is processed in layers to obtain multiple sub-selected areas, each of which corresponds to a color interval; A selection operator is generated using the content included in the sub-selection area, where the selection operator includes pixel points and blank points.

7. A remote sensing image target recognition device based on artificial intelligence, characterized in that: include: A preprocessing unit, for preprocessing the remote sensing image in response to the acquired remote sensing image to obtain a reference remote sensing image, wherein the preprocessing includes noise reduction, image correction and image cropping; A range extraction unit is used to perform range extraction in the reference remote sensing image to obtain a first region of interest, where the number of the first region of interest is at least one; An object extraction unit, configured to extract objects in the first region of interest to obtain at least one feature object; A judging unit, used to judge the completeness of the feature object and determine the missing position of the feature object; A retrieval unit, configured to create a second region of interest based on the missing position and perform a search in the second region of interest to find a covered object, wherein the type of the covered object is the same as the type of the covered feature object; Creating a second region of interest based on the missing location includes: Randomly select multiple feature points on the edge of the feature object; Calculate the eigenvalues ​​of the feature points, which include the color mean and the degree of color change in a certain direction; An attempt is made to reconstruct the missing position of the feature object according to the feature values ​​of the feature points and a second region of interest is created at the reconstructed missing position.

8. A remote sensing image target recognition system based on artificial intelligence, characterized in that: The system comprises: one or more memories for storing instructions; and One or more processors, configured to call and execute the instructions from the memory to perform the method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer readable storage medium comprises: The program, when the program is executed by a processor, the method according to any one of claims 1 to 6 is executed.

Images