Unmanned aerial vehicle image data enhancement method and system

Abstract

The embodiment of the invention provides an unmanned aerial vehicle image data enhancement method and system, and the method comprises the steps: obtaining an original fog image in response to a data enhancement request of an unmanned aerial vehicle image; the original fog image is used as input data, convolution training is carried out through a defogging neural network model, self-adaptive parameters of the original fog image are used as training features, defogged image data are output, and the self-adaptive parameters are jointly determined by transmissivity parameters and atmospheric light parameters; image data are converted into an HSV space, after image partitioning is carried out, pixel number redistribution is carried out on a histogram corresponding to each sub-block, and weighted fusion is carried out on pixel gray values according to a mapping function corresponding to the sub-block histogram; and dynamically enhancing the brightness component in the image data, and converting the image data in the HSV space back to the BRG space for corresponding storage and output.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a method and system for enhancing unmanned aerial vehicle (UAV) image data. Background Art

[0002] In recent years, with the rapid improvement of drone platform performance and breakthroughs in multispectral sensing technology, low-altitude remote sensing imagery has become an indispensable data source for fields such as digital city construction and precision agricultural and forestry management. However, due to factors such as atmospheric scattering and lighting conditions, raw imagery often suffers from quality issues such as low contrast and blurred details. These issues severely limit the accuracy of subsequent feature extraction and intelligent interpretation, especially when affected by fog. According to the Association for Unmanned Aerial Vehicle Systems (IDA), in 2023, nearly 70% of global low-altitude remote sensing data collected was affected by atmospheric scattering and uneven lighting, resulting in an effective information utilization rate of less than 45%, especially in areas prone to cloud and fog.

[0003] In summary, existing technologies for addressing these issues are typically based on atmospheric scattering models, which present certain drawbacks. First, the model parameter estimation accuracy is insufficient. The core of existing atmospheric models is to accurately assess light transmittance and atmospheric light. However, in drone remote sensing environments, the unique characteristics of complex objects at the depth and elevation levels lead to insufficient transmittance accuracy. Complex environments also make it difficult to obtain accurate and precise atmospheric light values. Second, detail recovery is difficult, making it impossible to effectively distinguish between details and noise in high-resolution images, which can easily lead to the smoothing of high-frequency information. Summary of the Invention

[0004] In response to the problems existing in the prior art, embodiments of the present invention provide a method and system for enhancing drone image data.

[0005] An embodiment of the present invention provides a method for enhancing drone image data, the method comprising:

[0006] Responding to data augmentation requests for drone images, obtaining the original fog image;

[0007] The original fog image is used as input data, convolution training is performed through a defogging neural network model, adaptive parameters of the original fog image are used as training features, and defogging image data is output, wherein the adaptive parameters are jointly determined by transmittance parameters and atmospheric light parameters;

[0008] The image data is converted into HSV space, and after image block division, the number of pixels of the histogram corresponding to each sub-block is redistributed, and the pixel grayscale values ​​are weightedly fused according to the mapping function corresponding to the sub-block histogram;

[0009] The brightness component in the image data is dynamically enhanced, and the image data in the HSV space is converted back to the BRG space for corresponding storage and output.

[0010] In one embodiment, the defogging neural network model includes:

[0011]

[0012] in, is the image data after defogging, is the adaptive parameter, is the original fog image, b is the constant bias;

[0013] The calculation formula of the adaptive parameter includes:

[0014]

[0015] Where A is the atmospheric light parameter, is the transmittance parameter.

[0016] In one embodiment, the method further comprises:

[0017] A dynamic threshold of the defogging neural network model is preset, the model output image is filtered by the dynamic threshold, and the output image data is linearly normalized.

[0018] In one embodiment, the method further comprises:

[0019] Counting the pixel intensity distribution of each sub-block, and establishing a grayscale histogram based on the pixel intensity distribution;

[0020] The number of pixels in the grayscale histogram is redistributed based on a preset contrast threshold, and the pixels in the sub-block exceeding the preset contrast threshold are distributed to sub-blocks corresponding to other grayscale histograms.

[0021] In one embodiment, the method further comprises:

[0022] Obtaining a mapping function of adjacent sub-blocks of pixels in the image data, where the mapping function is generated by histogram equalization;

[0023] A weight coefficient is set based on the distance from the adjacent sub-block to the pixel, and the grayscale value of the pixel is calculated in combination with the mapping function of the adjacent sub-block.

[0024] In one embodiment, the method further comprises:

[0025] In the HSV space, a Sigmoid function is performed on the brightness component of the image data.

[0026] An embodiment of the present invention provides a drone image data enhancement system, the system comprising:

[0027] an acquisition module, configured to obtain an original fog image in response to a data enhancement request for a UAV image;

[0028] a training module, configured to use the original fog image as input data, perform convolution training through a defogging neural network model, use adaptive parameters of the original fog image as training features, and output defogged image data, wherein the adaptive parameters are jointly determined by transmittance parameters and atmospheric light parameters;

[0029] A histogram module is used to convert the image data into HSV space, divide the image into blocks, redistribute the number of pixels in the histogram corresponding to each sub-block, and perform weighted fusion of pixel grayscale values ​​according to the mapping function corresponding to the sub-block histogram;

[0030] The enhancement module is used to dynamically enhance the brightness component in the image data, and convert the image data in the HSV space back to the BRG space for corresponding storage and output.

[0031] In one embodiment, the system further comprises:

[0032] A statistical module, configured to calculate the pixel intensity distribution of each sub-block and establish a grayscale histogram based on the pixel intensity distribution;

[0033] The redistribution module is used to redistribute the number of pixels in the grayscale histogram based on a preset contrast threshold as a redistribution basis, and the redistribution allocates the pixels in the sub-block exceeding the preset contrast threshold to the sub-blocks corresponding to other grayscale histograms.

[0034] An embodiment of the present invention provides an electronic device, including a processor and a memory;

[0035] The processor is connected to the memory;

[0036] The memory is used to store executable program code;

[0037] The processor reads the executable program code stored in the memory to run a program corresponding to the executable program code, so as to execute the method described in one or more embodiments.

[0038] An embodiment of the present invention provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the above-mentioned drone image data enhancement method are implemented.

[0039] In view of the above, in one or more embodiments of the present specification, in response to a data enhancement request for a drone image, an original fog image is obtained; the original fog image is used as input data, convolution training is performed through a defogging neural network model, and the adaptive parameters of the original fog image are used as training features to output defogged image data, where the adaptive parameters are jointly determined by the transmittance parameters and the atmospheric light parameters; the image data is converted to the HSV space, and after image block division, the number of pixels of the histogram corresponding to each sub-block is redistributed, and the pixel grayscale values ​​are weighted fused according to the mapping function corresponding to the sub-block histogram; the brightness component in the image data is dynamically enhanced, and the image data in the HSV space is converted back to the BRG space for corresponding storage and output. In this way, deep learning defogging and adaptive color correction can be combined through the enhancement framework of the traditional defogging model, and independent optimization of brightness and color can be achieved through the HSV space decoupling enhancement strategy; local histogram equalization and global nonlinear mapping are combined to take into account both detail enhancement and dynamic range control, thereby achieving a better defogging effect for the original fog image, thereby obtaining more accurate image data information while retaining image details. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0041] Figure 1 This is a flowchart of a method for enhancing drone image data provided by one embodiment of this specification.

[0042] Figure 2 An embodiment of this specification provides an image convolution training structure of an end-to-end network through the AOD-Net algorithm.

[0043] Figure 3 This is a flowchart of performing contrast enhancement and local enhancement on image data in the HSV color space, provided by an embodiment of this specification.

[0044] Figure 4 It is a quantitative indicator of a test image provided in an embodiment of this specification.

[0045] Figure 5 This is a comparison diagram of a test image before and after image enhancement provided by an embodiment of this specification.

[0046] Figure 6 This is a structural diagram of a drone image data enhancement system provided by an embodiment of this specification.

[0047] Figure 7 This is a structural diagram of an electronic device provided by an embodiment of this specification. DETAILED DESCRIPTION

[0048] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and are not intended to limit the scope of protection, applicability, or examples set forth in the claims. The functions and arrangements of the elements discussed may be changed without departing from the scope of protection of this specification. Various examples may omit, replace, or add various processes or components as needed. For example, the described method may be performed in an order different from the order described, and various steps may be added, omitted, or combined. In addition, features described relative to some examples may also be combined in other examples.

[0049] As used herein, the term "including" and its variations are open terms meaning "including but not limited to". The term "based on" means "based at least in part on". The terms "one embodiment" and "an embodiment" mean "at least one embodiment". The term "another embodiment" means "at least one other embodiment". The terms "first", "second", etc. may refer to different or the same objects. Other definitions may be included below, whether explicit or implicit. Unless the context clearly indicates otherwise, the definition of a term is consistent throughout the specification.

[0050] like Figure 1 As shown, an embodiment of the present invention provides a method for enhancing drone image data, comprising:

[0051] Step S102: In response to the data enhancement request of the drone image, an original fog image is obtained.

[0052] Specifically, in response to an image data enhancement request for an original drone image, a corresponding original fog image is obtained, wherein the original fog image includes an original image with poor image quality and blur due to interference from factors such as haze.

[0053] In step S104, the original fog image is used as input data, convolution training is performed through the defogging neural network model, and the adaptive parameters of the original fog image are used as training features to output the defogged image data, where the adaptive parameters are jointly determined by the transmittance parameters and the atmospheric light parameters.

[0054] Specifically, the defogging network model in this embodiment is an atmospheric scattering model improved by the ADO-Net algorithm, with adaptive parameters as the learning depth features during the training process. The formula of the traditional atmospheric scattering model is as follows:

[0055]

[0056] in, is the original fog map, is the image data after defogging, is the transmittance parameter, and A is the atmospheric light parameter.

[0057] In the traditional atmospheric scattering model, the formula for the transmittance parameter is:

[0058]

[0059] Among them, the transmittance parameter It is determined by the atmospheric scattering coefficient β and the scene depth d(x, y).

[0060] The above-mentioned traditional atmospheric scattering model needs to estimate the transmittance parameters (which depend on the scene depth) and atmospheric light parameters (which are easily affected by other factors, such as interference from bright ground objects) separately. The error will be amplified by the trigger operation.

[0061] In this embodiment, in order to overcome the problem of sensitivity to physical parameter estimation errors in the traditional atmospheric scattering model in the above steps, an adaptive parameter K(x) is introduced through the ADO-Net algorithm to determine the defogging neural network model in this embodiment, and train the original model. The defogging neural network model includes:

[0062]

[0063] in, is the image data after defogging, is the adaptive parameter, is the original fog image, b is a constant bias, which can be set to 1;

[0064] The calculation formula of the adaptive parameter includes:

[0065]

[0066] Where A is the atmospheric light parameter, is the transmittance parameter.

[0067] By combining A with Coupled to a single parameter, it solves the failure problem of traditional methods due to errors in dense fog / non-uniform fog. Compared with the traditional method's step-by-step estimation strategy of transmittance and atmospheric light, through A and The joint modeling mechanism can more effectively handle the impact mechanism of haze under complex atmospheric conditions. AOD-Net adopts an end-to-end network structure (structure diagram as shown in the figure). Figure 2As shown in Figure 3, the foggy image is directly mapped to the output fog-free image through multi-layer convolution training.

[0068] Furthermore, in order to avoid the possible pixel value overflow problem during the model training process, that is, the dehazing process, that is, the transmittance parameter has extreme positive values ​​(overexposure) or negative values ​​(underexposure), the model output image can be filtered through dynamic thresholds, and the output image data can be linearly normalized. The specific adjustment process can include adjusting through a dual-threshold dynamic range (upper threshold, lower threshold): first, the model output is truncated through dual thresholds, and the 1% and 99% quantiles of the processing results are used as the truncation thresholds low and high, and then a reasonable mapping of the pixel value range is achieved through linear normalization:

[0069] The formula for linear normalization includes:

[0070]

[0071] is the pixel value finally output by the model, high and low are the upper and lower thresholds respectively. The algorithm eliminates small extreme values ​​for the current pixel value, preserving details in dark and bright areas. It also linearly maps valid pixels to [0, 255] to maximize information utilization. Statistically adaptive threshold control effectively enhances dark area details while suppressing oversaturation in bright areas. In summary, the improved algorithm maintains the geometric texture characteristics of the image while exhibiting robustness to light fog, dense fog, and non-uniform haze.

[0072] Step S106 , converting the image data into HSV space, dividing the image into blocks, redistributing the number of pixels of the histogram corresponding to each sub-block, and performing weighted fusion on the pixel grayscale values ​​according to the mapping function corresponding to the sub-block histogram.

[0073] Specifically, the dehazed image data is converted to the HSV (Hue, Saturation, Value) space, and the image information is decoupled through the hue (H), saturation (S), and brightness (V) spaces to prevent the subsequent brightness (V) operations from affecting the hue (H) and saturation (S), thereby preventing color distortion. Contrast enhancement and local enhancement are then performed on the image data in the HSV color space. The steps include:

[0074] Step S302: Image block processing. The image data is divided into several non-overlapping rectangular sub-blocks (for example, an 8×8 grid). Each sub-block is processed independently. The choice of block size is a trade-off between computational efficiency and preservation of local features. Smaller sub-blocks can capture finer local features but increase computational complexity. An 8×8 block strategy can effectively extract medium-scale texture features while maintaining computational efficiency.

[0075] Step S304: Calculate local histogram: For each sub-block, count its grayscale value and draw the corresponding histogram to quantify the local contrast distribution characteristics.

[0076] Step S306: Contrast clipping. To avoid noise amplification caused by histogram peaks in local areas, the histogram of each sub-block is clipped. The number of pixels exceeding the preset contrast threshold is redistributed across all grayscale levels, thereby limiting the slope of the CDF (cumulative distribution function) and, in turn, controlling the magnitude of contrast enhancement.

[0077] Among them, the formula for CLAHE contrast limit includes:

[0078]

[0079] in, = N / L, where N is the number of pixels in the sub-block and L is the number of grayscale levels. A higher contrast threshold (number of pixels) results in more pronounced contrast enhancement, but at the risk of increased noise. A larger sub-block pixel size enhances local detail but increases computational complexity. Excessive contrast is evenly distributed across all grayscale levels to ensure balanced overall contrast.

[0080] Step S308: Histogram equalization: The clipped histogram is equalized to generate a mapping function to redistribute local pixel values ​​to a wider dynamic range.

[0081] Step S310: Multi-scale interpolation fusion. To avoid artifacts at sub-block boundaries, a bilinear interpolation algorithm is used to perform weighted fusion on the mapping results of adjacent sub-blocks. Specifically, for any pixel in the image, its final grayscale value is determined by the mapping functions of its adjacent sub-blocks, such as four adjacent sub-blocks. The weight coefficient is calculated by inversely proportionally calculating the distance from the pixel to the center of each sub-block. This multi-scale fusion mechanism maintains local enhancement while achieving smooth transitions between sub-blocks.

[0082] In the above steps, compared with the traditional global histogram equalization, the CLAHE method in this embodiment improves the spatial adaptability of the enhancement effect through local histogram processing; the contrast limiting mechanism improves the noise suppression efficiency; and the bilinear interpolation strategy reduces the incidence of boundary artifacts.

[0083] Step S108 , dynamically enhance the brightness component of the image data, and convert the image data in the HSV space back to the BRG space for corresponding storage and output.

[0084] Specifically, the brightness (V) component of the image data in the HSV space is dynamically enhanced to selectively increase the brightness of dark areas to address the problem of uneven illumination, complementing the CLAHE method in the above steps to improve the visibility of shadow areas. The dynamic enhancement method can fine-tune the dynamic range of the brightness component (V channel) in the HSV color space through a nonlinear enhancement mechanism based on the Sigmoid function. The core of the enhancement process is to achieve adaptive mapping of the brightness component through the Sigmoid function with saturation characteristics. Its mathematical formula is:

[0085]

[0086] Wherein, g is the gain coefficient and c is the cutoff threshold. The nonlinear enhancement mechanism of the Sigmoid function in this embodiment has obvious enhancement characteristics compared with traditional linear stretching or gamma correction. The gain coefficient g is used to control the steepness of the curve, and the cutoff threshold can control the center position of the enhancement. When it is greater than the threshold, the slope of the curve increases sharply, and the pixels in the dark area are significantly brightened. When it is lower than the threshold, the curve area is flat to prevent overexposure of the highlight area. In a specific implementation, adaptive adjustment of the dynamic range can be achieved by adjusting g and c. The saturation characteristic of Sigmoid can nonlinearly map the input dynamic range to the target interval, retaining more effective information. For the dynamically enhanced image data, it is converted to BGR space, and the usability of the result is ensured by reorganizing the color space and outputting it.

[0087] In this embodiment, the image enhancement method of this embodiment significantly improves image quality by integrating histogram equalization and shadow brightening algorithms based on ADO-Net dehazing. The enhanced results were compared with the original images, and four test images were analyzed using three quantitative metrics: PSNR, SSIM, and SAM. PSNR (Peak Signal-to-Noise Ratio) measures the pixel-level error between the enhanced image and the reference image (original image). It is calculated based on the mean square error (MSE). Generally, a value less than 20dB indicates significant distortion, 20-30dB is moderate, and greater than 30dB is considered near-lossless. SSIM (Structural Similarity Index) measures the similarity between two images across three dimensions: brightness, contrast, and structure, consistent with human visual perception. Values ​​between 0-0.6 indicate significant structural differences, between 0.6-0.8 indicate localized distortion, and values ​​greater than 0.8 are considered excellent. SAM (Spectral Angle Mapper) measures the pixel spectral fidelity by the angle between spectral vectors and is commonly used in remote sensing image analysis. When the value is less than 5, it is considered very good, and the medium performance range is 5-10. The test pictures in this embodiment use a total of 6 images. The specific quantitative indicators can be as follows Figure 4 As shown in the figure, the six control groups of the six images before and after image enhancement are as follows Figure 5 As shown, the upper picture in each group is the original image, and the lower picture is the enhanced image.

[0088] An embodiment of the present invention provides a method for enhancing drone image data. In response to a data enhancement request for a drone image, the method obtains an original fog image. Using the original fog image as input data, the method performs convolution training on a defogging neural network model, using the adaptive parameters of the original fog image as training features, and outputs defogged image data. The adaptive parameters are jointly determined by transmittance parameters and atmospheric light parameters. The method converts the image data into HSV space, partitions the image, redistributes the number of pixels in the histogram corresponding to each sub-block, and weightedly fuses the pixel grayscale values ​​based on the mapping function corresponding to the sub-block histogram. The method dynamically enhances the brightness component in the image data, and converts the image data in the HSV space back to the corresponding BRG space for storage and output. This method combines deep learning defogging with adaptive color correction through the enhancement framework of a traditional defogging model, and achieves independent optimization of brightness and color through an HSV space decoupling enhancement strategy. The method combines local histogram equalization with global nonlinear mapping to balance detail enhancement and dynamic range control, thereby achieving a better defogging effect for the original fog image.

[0089] See Figure 6 , Figure 6 This is a schematic diagram of the structure of a drone image data enhancement system provided by an embodiment of the present application. Figure 6 As shown, the system includes:

[0090] An acquisition module S602 is configured to acquire an original fog image in response to a data enhancement request for a drone image;

[0091] A training module S604 is configured to perform convolution training on a defogging neural network model using the original fog image as input data, using adaptive parameters of the original fog image as training features, and outputting defogged image data, wherein the adaptive parameters are jointly determined by transmittance parameters and atmospheric light parameters;

[0092] The histogram module S606 is used to convert the image data into HSV space, divide the image into blocks, redistribute the number of pixels in the histogram corresponding to each sub-block, and perform weighted fusion of the pixel grayscale values ​​according to the mapping function corresponding to the sub-block histogram;

[0093] The enhancement module S608 is used to dynamically enhance the brightness component of the image data, and convert the image data in the HSV space back to the BRG space for corresponding storage and output. In another embodiment, a drone image data enhancement system further includes:

[0094] In another embodiment, a drone image data enhancement system further includes:

[0095] A statistical module, configured to calculate the pixel intensity distribution of each sub-block and establish a grayscale histogram based on the pixel intensity distribution;

[0096] The redistribution module is used to redistribute the number of pixels in the grayscale histogram based on a preset contrast threshold as a redistribution basis, and the redistribution allocates the pixels in the sub-block exceeding the preset contrast threshold to the sub-blocks corresponding to other grayscale histograms.

[0097] Those skilled in the art will clearly understand that the technical solutions of the embodiments of the present application can be implemented with the help of software and / or hardware. "Unit" and "module" in this specification refer to software and / or hardware that can independently perform or cooperate with other components to perform specific functions, where the hardware can be, for example, a field-programmable gate array (FPGA) or an integrated circuit (IC).

[0098] Each processing unit and / or module in the embodiments of the present application may be implemented by an analog circuit that implements the functions described in the embodiments of the present application, or may be implemented by software that executes the functions described in the embodiments of the present application.

[0099] See also Figure 7 , which shows a schematic diagram of the structure of an electronic device involved in an embodiment of the present application, the electronic device can be used to implement Figure 1 The method in the embodiment shown. Figure 7 As shown, the electronic device 700 may include: at least one processor 701 , at least one network interface 704 , a user interface 703 , a memory 705 , and at least one communication bus 702 .

[0100] The communication bus 702 is used to implement the connection and communication between these components.

[0101] The user interface 703 may include a display screen (Display) and a camera (Camera). Optionally, the user interface 703 may also include a standard wired interface and a wireless interface.

[0102] The network interface 704 may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface).

[0103] The processor 701 may include one or more processing cores. The processor 701 utilizes various interfaces and circuits to connect various components within the electronic device 700. It executes instructions, programs, code sets, or instruction sets stored in the memory 705 and accesses data stored in the memory 705 to perform various functions and process data within the electronic device 700. Optionally, the processor 701 may be implemented in hardware using at least one of a digital signal processing (DSP), a field-programmable gate array (FPGA), and a programmable logic array (PLA). The processor 701 may integrate one or a combination of a central processing unit (CPU), a graphics processing unit (GPU), and a modem. The CPU primarily processes the operating system, user interface, and application programs; the GPU is responsible for rendering and drawing content displayed on the display; and the modem handles wireless communications. It is understood that the modem may also be implemented independently of the processor 701 and implemented on a separate chip.

[0104] Among them, the memory 705 may include a random access memory (RAM) or a read-only memory (Read-Only Memory). Optionally, the memory 705 includes a non-transitory computer-readable storage medium. The memory 705 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 705 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area may store data involved in the above-mentioned various method embodiments, etc. The memory 705 may also be optionally at least one storage device located away from the aforementioned processor 701. As Figure 7 As shown, the memory 705 as a computer storage medium may include an operating system, a network communication module, a user interface module, and program instructions.

[0105] exist Figure 7 In the electronic device 700 shown, the user interface 703 is mainly used to provide an input interface for the user and obtain the data input by the user; and the processor 701 can be used to call the interactive application based on image generation stored in the memory 705, and perform the following operations: in response to the data enhancement request of the drone image, obtain the original fog image; use the original fog image as input data, perform convolution training through the defogging neural network model, use the adaptive parameters of the original fog image as training features, and output the defogged image data, where the adaptive parameters are jointly determined by the transmittance parameters and the atmospheric light parameters; convert the image data into the HSV space, divide the image into blocks, redistribute the number of pixels of the histogram corresponding to each sub-block, and perform weighted fusion on the pixel grayscale values ​​according to the mapping function corresponding to the sub-block histogram; dynamically enhance the brightness component in the image data, and convert the image data in the HSV space back to the BRG space for corresponding storage and output.

[0106] The present application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above method. The computer-readable storage medium may include, but is not limited to, any type of disk, including a floppy disk, an optical disk, a DVD, a CD-ROM, a microdrive, a magneto-optical disk, a ROM, a RAM, an EPROM, an EEPROM, a DRAM, a VRAM, a flash memory device, a magnetic card or an optical card, a nanosystem (including a molecular memory IC), or any type of medium or device suitable for storing instructions and / or data.

[0107] It should be noted that for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that this application is not limited by the order of the actions described, because according to this application, 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 the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by this application.

[0108] 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.

[0109] In the several embodiments provided in this application, it should be understood that the disclosed devices can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, 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 service interface, and the indirect coupling or communication connection of the device or unit can be electrical or other forms.

[0110] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0111] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0112] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a memory and includes a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned memory includes: U disk, read-only memory (ROM), random access memory (RAM), mobile hard disk, magnetic disk, or optical disk, etc., various media that can store program code.

[0113] Those skilled in the art will appreciate that all or part of the steps in the various methods of the above embodiments may be completed by instructing related hardware through a program. The program may be stored in a computer-readable memory, which may include a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0114] The foregoing description of this specification describes specific embodiments. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different from that described in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order shown or the sequential order to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A method for enhancing drone image data, characterized in that: The method comprises: Responding to data augmentation requests for drone images, obtaining the original fog image; The original fog image is used as input data, convolution training is performed through a defogging neural network model, adaptive parameters of the original fog image are used as training features, and defogging image data is output, wherein the adaptive parameters are jointly determined by transmittance parameters and atmospheric light parameters; The image data is converted into HSV space, and after image block division, the number of pixels of the histogram corresponding to each sub-block is redistributed, and the pixel grayscale values ​​are weightedly fused according to the mapping function corresponding to the sub-block histogram; The brightness component in the image data is dynamically enhanced, and the image data in the HSV space is converted back to the BRG space for corresponding storage and output.

2. The method according to claim 1, characterized in that The defogging neural network model includes: ， in, is the image data after defogging, is the adaptive parameter, is the original fog image, b is the constant bias; The calculation formula of the adaptive parameter includes: ， Where A is the atmospheric light parameter, is the transmittance parameter.

3. The method according to claim 2, characterized in that The convolution training process also includes: A dynamic threshold of the defogging neural network model is preset, the model output image is filtered by the dynamic threshold, and the output image data is linearly normalized.

4. The method according to claim 1, wherein The redistributing the number of pixels of the histogram corresponding to each sub-block includes: Counting the pixel intensity distribution of each sub-block, and establishing a grayscale histogram based on the pixel intensity distribution; The number of pixels in the grayscale histogram is redistributed based on a preset contrast threshold, and the pixels in the sub-block exceeding the preset contrast threshold are distributed to sub-blocks corresponding to other grayscale histograms.

5. The method according to claim 4, characterized in that The weighted fusion of pixel grayscale values ​​according to the mapping function corresponding to the sub-block histogram includes: Obtaining a mapping function of adjacent sub-blocks of pixels in the image data, where the mapping function is generated by histogram equalization; A weight coefficient is set based on the distance from the adjacent sub-block to the pixel, and the grayscale value of the pixel is calculated in combination with the mapping function of the adjacent sub-block.

6. The method according to claim 1, characterized in that The dynamically enhancing the brightness component in the image data comprises: In the HSV space, a Sigmoid function is performed on the brightness component of the image data.

7. A system for repairing sparse cloud pollution in optical remote sensing images, characterized in that: The system comprises: an acquisition module, configured to obtain an original fog image in response to a data enhancement request for a drone image; a training module, configured to use the original fog image as input data, perform convolution training through a defogging neural network model, use adaptive parameters of the original fog image as training features, and output defogged image data, wherein the adaptive parameters are jointly determined by transmittance parameters and atmospheric light parameters; A histogram module is used to convert the image data into HSV space, divide the image into blocks, redistribute the number of pixels in the histogram corresponding to each sub-block, and perform weighted fusion of pixel grayscale values ​​according to the mapping function corresponding to the sub-block histogram; The enhancement module is used to dynamically enhance the brightness component in the image data, and convert the image data in the HSV space back to the BRG space for corresponding storage and output.

8. The system according to claim 7, characterized in that The system further comprises: A statistical module, configured to calculate the pixel intensity distribution of each sub-block and establish a grayscale histogram based on the pixel intensity distribution; The redistribution module is used to redistribute the number of pixels in the grayscale histogram based on a preset contrast threshold as a redistribution basis, and the redistribution allocates the pixels in the sub-block exceeding the preset contrast threshold to the sub-blocks corresponding to other grayscale histograms.

9. An electronic device, characterized in that: including a processor and a memory; The processor is connected to the memory; The memory is used to store executable program code; The processor reads the executable program code stored in the memory to run a program corresponding to the executable program code, so as to execute the method according to any one of claims 1 to 5.

10. A computer-readable storage medium, characterized in that A computer program is stored thereon, and when the computer program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

Images