A UAV image defogging system and method based on artificial intelligence

Abstract

The present invention relates to the technical field of unmanned aerial vehicle image processing, and specifically to an unmanned aerial vehicle image defogging system and method based on artificial intelligence, comprising analyzing the similarity between historical unmanned aerial vehicle foggy images and unmanned aerial vehicle foggy images in terms of the impact of haze on images, and obtaining a similarity value of haze impact on characteristic images; analyzing the image similarity between historical unmanned aerial vehicle foggy images and unmanned aerial vehicle foggy images; analyzing the overall similarity between historical unmanned aerial vehicle foggy images and unmanned aerial vehicle foggy images, and obtaining characteristic historical unmanned aerial vehicle foggy images; obtaining characteristic historical unmanned aerial vehicle foggy images corresponding to unmanned aerial vehicle foggy images in a current cycle, obtaining historical unmanned aerial vehicle shooting records to which the characteristic historical unmanned aerial vehicle foggy images belong, and optimizing unmanned aerial vehicle foggy images shot by unmanned aerial vehicles in the current cycle based on the corresponding historical unmanned aerial vehicle shooting records, and performing intelligent management on unmanned aerial vehicle image defogging.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle image processing, and in particular to an unmanned aerial vehicle image defogging system and method based on artificial intelligence. Background Art

[0002] When a drone is flying and shooting, once it encounters foggy weather or smoke, the quality of the images taken by the drone will be reduced, details will be lost, and other problems will occur. In addition, defogging of drone images is a very complex image processing problem. Traditional image processing methods often cannot meet the requirements of real-time, high efficiency, and high quality. At this time, it is necessary to use artificial intelligence technology in the drone image defogging system. The drone defogging system using artificial intelligence technology includes but is not limited to the following advantages: 1. Improve the image defogging effect. Artificial intelligence technology can learn a large number of fog-free image samples through deep learning and other methods, so as to restore the fog-free image more accurately and restore a clearer and more detailed image; 2. Adaptability. Artificial intelligence technology has adaptive learning ability and can adapt to the image defogging requirements under different environmental conditions. It can adjust the parameters of the defogging-related algorithm in real time according to the different image shooting conditions, and improve the adaptability to drone image defogging; 3. Reduce cost consumption. Traditional image processing methods usually require a lot of manual intervention and parameter adjustment. Usually, this process requires a lot of time and human resources. Artificial intelligence technology can realize automatic processing to a certain extent, reducing labor costs and greatly improving efficiency.

[0003] When drones are taking aerial photos, the density of haze in the environment is often uneven, with changes in regional concentration. In addition, a considerable number of defogging algorithms only perform defogging on images based on characteristic scenes or common assumptions, and have poor adaptability to other scenes. They can only process certain environmental conditions of drone images. Once the environment of the processed drone images does not meet the assumed environmental conditions, it is very likely that the color, details or edges of the defogging drone images will be distorted or have artifacts, reducing the authenticity and accuracy of the images. Summary of the invention

[0004] The purpose of the present invention is to provide an artificial intelligence-based drone image defogging system and method to solve the problems existing in the above-mentioned background technology.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: a method for defogging drone images based on artificial intelligence, the method comprising:

[0006] Step S100: construct a drone image defogging cloud platform, obtain drone foggy images taken by drones in the current period, obtain historical drone shooting records, obtain historical drone foggy images from historical drone shooting records, analyze the similarity between the historical drone foggy images and the drone foggy images in terms of the impact of haze on the images, and obtain the similarity value of haze impact on the feature images;

[0007] Step S200: Based on the historical drone foggy images in the acquired historical drone shooting records, the image similarity between the historical drone foggy images and the drone foggy images is analyzed to obtain a feature image similarity value;

[0008] Step S300: Analyze the overall similarity between the historical drone foggy image and the drone foggy image based on the similarity value of the characteristic image haze effect between the historical drone foggy image and the drone foggy image, and obtain the characteristic historical drone foggy image;

[0009] Step S400: Obtain the characteristic historical drone foggy image corresponding to the drone foggy image in the current period, obtain the historical drone shooting record to which the characteristic historical drone foggy image belongs, optimize the drone foggy image shot by the drone in the current period, and perform intelligent management of drone image defogging.

[0010] Furthermore, step S100 includes:

[0011] Step S101: constructing a drone defogging cloud platform, using a drone to photograph the area with haze in the current period, and obtaining a drone foggy image in the current period;

[0012] Step S102: using drones to photograph historical areas with smog and historical areas without smog during the historical period, and collecting the photographed images to obtain historical drone photography records;

[0013] Step S103: Acquire each historical drone shooting record stored in the cloud platform, and acquire historical drone foggy images from the historical drone shooting records;

[0014] Step S104: analyzing the similarity between the drone foggy image and several historical drone foggy images in each historical drone shooting record in terms of the impact of fog on the image, wherein the specific process of analyzing the similarity between the drone foggy image and the a-th historical drone foggy image in the c-th historical drone shooting record in terms of the impact of fog on the image is to adjust the image size of the a-th historical drone foggy image based on the drone foggy image, perform image preprocessing on the a-th historical drone foggy image and the drone foggy image, and perform grayscale conversion on the drone foggy image and the a-th historical drone foggy image;

[0015] Step S105: Obtain the dark channel image corresponding to the drone foggy image and the ath historical drone foggy image, wherein the dark channel image corresponding to the ath historical drone foggy image :

[0016] ;

[0017] in, The image window of pixel x in the a-th historical UAV foggy image is preset; represents the gray value of pixel y in the a-th historical UAV foggy image;

[0018] Step S106: Calculate the characteristic haze impact value of each pixel point in the a-th historical drone foggy image and the drone foggy image, where the characteristic haze impact value of the pixel point x in the a-th historical drone foggy image is :

[0019] ;

[0020] in, , ; is the preset characteristic haze constant; It is represented as the gray value of pixel x in the a-th historical UAV foggy image;

[0021] in, The value usually ranges from 0.95 to 1;

[0022] Step S107: Calculate the similarity value of the characteristic haze impact between the drone foggy image and the ath historical drone foggy image :

[0023] ;

[0024] in, is the total number of pixels in the foggy image of the drone; is the characteristic haze impact value of the i-th pixel in the foggy image of the drone; is the characteristic haze impact value of the i-th pixel in the a-th historical UAV foggy image.

[0025] Further, step S200 includes:

[0026] Step S201: acquiring each historical drone shooting record in the cloud platform, respectively acquiring a historical drone fog image of a historical area shot by a drone from each historical drone shooting record, and acquiring a drone fog image of a current period;

[0027] Step S202: analyzing the image similarity between the drone foggy image and several historical drone foggy images in each historical drone shooting record, wherein the specific analysis process of the image similarity between the drone foggy image and the f-th historical drone foggy image in the e-th historical drone shooting record is to adjust the image size of the f-th historical drone foggy image according to the drone foggy image, perform image preprocessing on the adjusted f-th historical drone foggy image and the drone foggy image, and perform image grayscale conversion on the preprocessed f-th historical drone foggy image and the drone foggy image;

[0028] Step S203: Calculate the brightness similarity value of the marked image between the drone foggy image and the fth historical drone foggy image :

[0029] ;

[0030] in, is the average gray value of each pixel in the f-th historical UAV foggy image; is the average gray value of each pixel in the foggy image of the drone; is a preset first labeled image proximity constant;

[0031] Step S204: Calculate the similarity value of the contrast between the marked image of the drone foggy image and the fth historical drone foggy image :

[0032] ;

[0033] in, is the covariance between the UAV foggy image and the fth historical UAV foggy image; is the variance of the f-th historical UAV foggy image; is the variance of the foggy UAV image; is a preset second marker image similarity constant;

[0034] Step S205: Calculate the similarity value of the labeled image structure between the drone foggy image and the fth historical drone foggy image :

[0035] ;

[0036] in, is the covariance of the f-th historical drone foggy image; is the covariance of the foggy drone image; is a preset third marker image similar constant;

[0037] Step S206: Calculate the similarity value of the feature image between the drone fog image and the f-th historical drone fog image .

[0038] Furthermore, step S300 includes:

[0039] Step S301: Obtain the drone foggy image in the current cycle, and the haze impact similarity value of the characteristic image of several historical drone foggy images in each historical drone shooting record, and obtain the maximum value of the haze impact similarity value of the drone foggy image and the characteristic image of several historical drone foggy images in each historical drone shooting record. and minimum value ;

[0040] Step S302: Obtain the drone fog image in the current period, and the feature image similarity value of several historical drone fog images in each historical drone shooting record, and obtain the maximum value of the drone fog image and the feature image similarity value of several historical drone fog images in each historical drone shooting record. and minimum value ;

[0041] Step S303: Analyze the overall similarity between the drone fog image and several historical drone fog images in each historical drone shooting record, where the drone fog image is similar to the first The first in the historic drone photography record The specific analysis process of the overall similarity between the historical drone foggy images is as follows: The first in the historic drone photography record The overall similarity of features between historical drone foggy images :

[0042] ;

[0043] in, For the foggy image of the drone, The first in the historic drone photography record The similarity value of feature images between historical UAV foggy images; For the foggy image of the drone, The first in the historic drone photography record The characteristic haze impact values ​​between the historical drone foggy images are similar;

[0044] Step S304: When the drone has foggy image, The first in the historic drone photography record If the overall similarity value of the features between the historical drone foggy images is greater than the preset overall similarity threshold of the features, the drone foggy image is judged to be The first in the historic drone photography record The historical drone foggy images are similar in overall quality. Historical drone foggy images, recorded as the characteristic historical drone foggy images;

[0045] In the above steps, the similarity value between the drone foggy image and the feature image of several historical drone foggy images in each historical drone shooting record indicates whether the drone foggy image is similar to the historical drone foggy image. However, image similarity does not mean that the haze conditions of the images are the same. Therefore, by combining the drone foggy image with the similarity value of the feature image haze influence of several historical drone foggy images in each historical drone shooting record, an overall judgment is made on the similarity between the historical drone foggy image and the drone foggy image in the current period, so that the judgment result is more accurate, and it is also more conducive to improving the defogging accuracy of the drone foggy image in the subsequent period.

[0046] Furthermore, step S400 includes:

[0047] Step S401: Acquire each characteristic historical drone foggy image of the drone foggy image in the current period, acquire the historical drone shooting record corresponding to each characteristic historical drone foggy image, and acquire the characteristic historical drone fog-free image of the historical area corresponding to the characteristic historical drone foggy image from the historical drone shooting record;

[0048] Step S402: Based on each characteristic historical drone foggy image of the drone foggy image, and the characteristic historical drone fog-free image corresponding to each characteristic historical drone foggy image, the drone foggy image in the current period is optimized after being defogged by the defogging algorithm, and various parameters of the algorithm are adjusted to perform intelligent management of drone image defogging.

[0049] In order to better implement the above method, a drone image defogging system is also proposed. The drone image defogging system includes a haze impact degree analysis module, an image similarity degree analysis module, a feature history drone foggy image module, and an image defogging intelligent management module.

[0050] The haze impact degree analysis module is used to analyze the similarity of haze impact on images between historical drone foggy images and drone foggy images, and obtain the similarity value of haze impact on feature images;

[0051] The image similarity analysis module is used to analyze the image similarity between the historical drone foggy image and the drone foggy image to obtain the feature image similarity value;

[0052] The characteristic historical drone foggy image module is used to analyze the overall similarity between the historical drone foggy images and the drone foggy images to obtain the characteristic historical drone foggy images;

[0053] The image defogging intelligent management module is used to obtain the characteristic historical drone foggy images corresponding to the drone foggy images in the current period, optimize the drone foggy images taken by the drone in the current period, and perform intelligent management of drone image defogging.

[0054] Furthermore, the haze impact degree analysis module includes a characteristic haze impact value unit and a characteristic haze impact similar value unit;

[0055] A characteristic haze impact value unit is used to calculate the characteristic haze impact value of each pixel point of the historical drone foggy image and the drone foggy image;

[0056] The characteristic haze impact similarity value unit is used to calculate the characteristic haze impact similarity value between the drone foggy image and the historical drone foggy image.

[0057] Furthermore, the image similarity analysis module includes an image similarity analysis unit and a feature image similarity value unit;

[0058] An image similarity analysis unit is used to analyze the image similarity between the drone foggy image and a plurality of historical drone foggy images in each historical drone shooting record;

[0059] The feature image similarity value unit is used to calculate the feature image similarity value between the drone foggy image and the historical drone foggy image.

[0060] Further, the feature historical drone foggy image module includes a feature overall similarity value unit and a feature historical drone foggy image unit;

[0061] A feature overall similarity value unit is used to calculate the feature overall similarity value between the drone foggy image and the historical drone foggy image in the historical drone shooting record;

[0062] The feature historical drone foggy image unit is used to obtain the feature historical drone foggy image of the drone foggy image according to the feature overall similarity value between the drone foggy image and the historical drone foggy image in the historical drone shooting record.

[0063] Further, the image defogging intelligent management module includes an image defogging intelligent management unit;

[0064] The image defogging intelligent management unit is used to optimize the images of the drone foggy images in the current period after being defogged by the defogging algorithm according to each feature historical drone foggy image of the drone foggy image and the feature historical drone fog-free image corresponding to each feature historical drone foggy image, and adjust various parameters of the algorithm to perform intelligent management of drone image defogging.

[0065] Compared with the prior art, the beneficial effects achieved by the present invention are: the present invention realizes intelligent management of drone image defogging, has strong adaptability to various scenes when defogging images, has no specific requirements for the environment in which the drone is photographed during the defogging of drone images, and uses a large number of historical drone images in the defogging of drone images, so as to more accurately restore the fog-free image and restore a clearer and more detailed image, and can also adjust the parameters of the defogging related algorithm in real time according to different image shooting conditions, thereby greatly improving the adaptability to drone image defogging. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] The accompanying drawings are used to provide a further understanding of the present invention and constitute part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0067] Figure 1 It is a method flow chart of an artificial intelligence-based UAV image defogging system and method of the present invention;

[0068] Figure 2 It is a module schematic diagram of an artificial intelligence-based UAV image defogging system and method of the present invention. DETAILED DESCRIPTION

[0069] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0070] See also Figure 1-Figure 2 The present invention provides a technical solution: a method for defogging drone images based on artificial intelligence, the method comprising:

[0071] Step S100: construct a drone image defogging cloud platform, obtain drone foggy images taken by drones in the current period, obtain historical drone shooting records, obtain historical drone foggy images from historical drone shooting records, analyze the similarity between the historical drone foggy images and the drone foggy images in terms of the impact of haze on the images, and obtain the similarity value of haze impact on the feature images;

[0072] Wherein, step S100 includes:

[0073] Step S101: constructing a drone defogging cloud platform, using a drone to photograph the area with haze in the current period, and obtaining a drone foggy image in the current period;

[0074] Step S102: using drones to photograph historical areas with smog and historical areas without smog during the historical period, and collecting the photographed images to obtain historical drone photography records;

[0075] Step S103: Acquire each historical drone shooting record stored in the cloud platform, and acquire historical drone foggy images from the historical drone shooting records;

[0076] Step S104: analyzing the similarity between the drone foggy image and several historical drone foggy images in each historical drone shooting record in terms of the impact of fog on the image, wherein the specific process of analyzing the similarity between the drone foggy image and the a-th historical drone foggy image in the c-th historical drone shooting record in terms of the impact of fog on the image is to adjust the image size of the a-th historical drone foggy image based on the drone foggy image, perform image preprocessing on the a-th historical drone foggy image and the drone foggy image, and perform grayscale conversion on the drone foggy image and the a-th historical drone foggy image;

[0077] For example, image preprocessing includes operations such as resizing the image, removing noise, and smoothing the image;

[0078] Step S105: Obtain the dark channel image corresponding to the drone foggy image and the ath historical drone foggy image, wherein the dark channel image corresponding to the ath historical drone foggy image :

[0079] ;

[0080] in, The image window of pixel x in the a-th historical UAV foggy image is preset; represents the gray value of pixel y in the a-th historical UAV foggy image;

[0081] Step S106: Calculate the characteristic haze impact value of each pixel point in the a-th historical drone foggy image and the drone foggy image, where the characteristic haze impact value of the pixel point x in the a-th historical drone foggy image is :

[0082] ;

[0083] in, , ; is the preset characteristic haze constant; It is represented as the gray value of pixel x in the a-th historical UAV foggy image;

[0084] Step S107: Calculate the similarity value of the characteristic haze impact between the drone foggy image and the ath historical drone foggy image :

[0085] ;

[0086] in, is the total number of pixels in the foggy image of the drone; is the characteristic haze impact value of the i-th pixel in the foggy image of the drone; is the characteristic haze impact value of the i-th pixel in the a-th historical UAV foggy image;

[0087] Step S200: Based on the historical drone foggy images in the acquired historical drone shooting records, the image similarity between the historical drone foggy images and the drone foggy images is analyzed to obtain a feature image similarity value;

[0088] Wherein, step S200 includes:

[0089] Step S201: acquiring each historical drone shooting record in the cloud platform, respectively acquiring a historical drone fog image of a historical area shot by a drone from each historical drone shooting record, and acquiring a drone fog image of a current period;

[0090] Step S202: analyzing the image similarity between the drone foggy image and several historical drone foggy images in each historical drone shooting record, wherein the specific analysis process of the image similarity between the drone foggy image and the f-th historical drone foggy image in the e-th historical drone shooting record is to adjust the image size of the f-th historical drone foggy image according to the drone foggy image, perform image preprocessing on the adjusted f-th historical drone foggy image and the drone foggy image, and perform image grayscale conversion on the preprocessed f-th historical drone foggy image and the drone foggy image;

[0091] Step S203: Calculate the brightness similarity value of the marked image between the drone foggy image and the fth historical drone foggy image :

[0092] ;

[0093] in, is the average gray value of each pixel in the f-th historical UAV foggy image; is the average gray value of each pixel in the foggy image of the drone; is a preset first labeled image approximation constant;

[0094] For example, the average gray value of each pixel in the second historical drone foggy image is 80, the average gray value of each pixel in the foggy image of the drone is 70; is 1;

[0095] Calculate the brightness similarity between the drone fog image and the second historical drone fog image ;

[0096] Step S204: Calculate the similarity value of the contrast between the marked image of the drone foggy image and the fth historical drone foggy image :

[0097] ;

[0098] in, is the covariance between the UAV foggy image and the fth historical UAV foggy image; is the variance of the f-th historical drone foggy image; is the variance of the foggy UAV image; is a preset second marker image similarity constant;

[0099] Step S205: Calculate the similarity value of the labeled image structure between the drone foggy image and the fth historical drone foggy image :

[0100] ;

[0101] in, is the covariance of the f-th historical drone foggy image; is the covariance of the foggy drone image; is a preset third marker image similar constant;

[0102] Step S206: Calculate the similarity value of the feature image between the drone fog image and the f-th historical drone fog image ;

[0103] Step S300: Analyze the overall similarity between the historical drone foggy image and the drone foggy image based on the similarity value of the characteristic image haze effect between the historical drone foggy image and the drone foggy image, and obtain the characteristic historical drone foggy image;

[0104] Wherein, step S300 includes:

[0105] Step S301: Obtain the drone foggy image in the current cycle, and the haze impact similarity value of the characteristic image of several historical drone foggy images in each historical drone shooting record, and obtain the maximum value of the haze impact similarity value of the drone foggy image and the characteristic image of several historical drone foggy images in each historical drone shooting record. and minimum value ;

[0106] Step S302: Obtain the drone fog image in the current period, and the feature image similarity value of several historical drone fog images in each historical drone shooting record, and obtain the maximum value of the drone fog image and the feature image similarity value of several historical drone fog images in each historical drone shooting record. and minimum value ;

[0107] Step S303: Analyze the overall similarity between the drone fog image and several historical drone fog images in each historical drone shooting record, where the drone fog image is similar to the first The first in the historic drone photography record The specific analysis process of the overall similarity between the historical drone foggy images is as follows: The first in the historic drone photography record The overall similarity of features between historical drone foggy images :

[0108] ;

[0109] in, For the foggy image of the drone, The first in the historic drone photography record The similarity value of feature images between historical UAV foggy images; For the foggy image of the drone, The first in the historic drone photography record The characteristic haze impact values ​​between the historical drone foggy images are similar;

[0110] For example, the maximum value of the characteristic image similarity between the drone fog image and several historical drone fog images in each historical drone shooting record and minimum value 0.90 and 0.50 respectively; the maximum value of the haze effect of the drone foggy image and the characteristic image of several historical drone foggy images in each historical drone shooting record and minimum value 100 and 10 respectively; UAV foggy image, and The first in the historic drone photography record The similarity value of the feature images between the historical UAV foggy images is 0.80; the drone fog image is The first in the historic drone photography record The similarity of the characteristic haze effect between the historical drone foggy images is 80;

[0111] Calculate the foggy image of the drone, and The first in the historic drone photography record The overall similarity of features between historical drone foggy images ;

[0112] Step S304: When the drone has foggy image, The first in the historic drone photography record If the overall similarity value of the features between the historical drone foggy images is greater than the preset overall similarity threshold of the features, the drone foggy image is judged to be The first in the historic drone photography record The historical drone foggy images are similar in overall quality. Historical drone foggy images, recorded as the characteristic historical drone foggy images;

[0113] Step S400: obtaining a characteristic historical drone foggy image corresponding to the drone foggy image in the current period, obtaining a historical drone shooting record to which the characteristic historical drone foggy image belongs, optimizing the drone foggy image shot by the drone in the current period, and performing intelligent management on drone image defogging;

[0114] Wherein, step S400 includes:

[0115] Step S401: Acquire each characteristic historical drone foggy image of the drone foggy image in the current period, acquire the historical drone shooting record corresponding to each characteristic historical drone foggy image, and acquire the characteristic historical drone fog-free image of the historical area corresponding to the characteristic historical drone foggy image from the historical drone shooting record;

[0116] Step S402: based on each characteristic historical drone foggy image of the drone foggy image and the characteristic historical drone fog-free image corresponding to each characteristic historical drone foggy image, the drone foggy image in the current period is optimized after being defogged by the defogging algorithm, and various parameters of the algorithm are adjusted to perform intelligent management of drone image defogging;

[0117] For example, various parameters include minimum filter window size, constrained smoothing parameters, etc.

[0118] In order to better implement the above method, a drone image defogging system is also proposed. The drone image defogging system includes a haze impact degree analysis module, an image similarity degree analysis module, a feature history drone foggy image module, and an image defogging intelligent management module.

[0119] The haze impact degree analysis module is used to analyze the similarity of haze impact on images between historical drone foggy images and drone foggy images, and obtain the similarity value of haze impact on feature images;

[0120] The image similarity analysis module is used to analyze the image similarity between the historical drone foggy image and the drone foggy image to obtain the feature image similarity value;

[0121] The characteristic historical drone foggy image module is used to analyze the overall similarity between the historical drone foggy images and the drone foggy images to obtain the characteristic historical drone foggy images;

[0122] The image defogging intelligent management module is used to obtain the characteristic historical drone foggy images corresponding to the drone foggy images in the current cycle, optimize the drone foggy images taken by the drone in the current cycle, and perform intelligent management of drone image defogging;

[0123] Among them, the haze impact degree analysis module includes characteristic haze impact value unit and characteristic haze impact similar value unit;

[0124] A characteristic haze impact value unit is used to calculate the characteristic haze impact value of each pixel point of the historical drone foggy image and the drone foggy image;

[0125] A characteristic haze impact similarity value unit is used to calculate the characteristic haze impact similarity value between the drone foggy image and the historical drone foggy image;

[0126] The image similarity analysis module includes an image similarity analysis unit and a feature image similarity value unit;

[0127] An image similarity analysis unit is used to analyze the image similarity between the drone foggy image and a plurality of historical drone foggy images in each historical drone shooting record;

[0128] A feature image similarity value unit is used to calculate the feature image similarity value between the drone foggy image and the historical drone foggy image;

[0129] Among them, the feature historical drone foggy image module includes a feature overall similarity value unit and a feature historical drone foggy image unit;

[0130] A feature overall similarity value unit is used to calculate the feature overall similarity value between the drone foggy image and the historical drone foggy image in the historical drone shooting record;

[0131] A characteristic historical drone foggy image unit is used to obtain a characteristic historical drone foggy image of the drone foggy image according to a characteristic overall similarity value between the drone foggy image and the historical drone foggy image in the historical drone shooting record;

[0132] Wherein, the image defogging intelligent management module includes an image defogging intelligent management unit;

[0133] The image defogging intelligent management unit is used to optimize the images of the drone foggy images in the current period after being defogged by the defogging algorithm according to each feature historical drone foggy image of the drone foggy image and the feature historical drone fog-free image corresponding to each feature historical drone foggy image, and adjust various parameters of the algorithm to perform intelligent management of drone image defogging.

[0134] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0135] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for defogging drone images based on artificial intelligence, characterized in that: The method comprises: Step S100: constructing a drone image defogging cloud platform, obtaining drone foggy images taken by drones in the current period, obtaining historical drone shooting records, obtaining historical drone foggy images from the historical drone shooting records, analyzing the similarity between the historical drone foggy images and the drone foggy images in terms of the impact of haze on the images, and obtaining a similar value of haze impact on the feature images; The step S100 includes: Step S101: constructing a drone defogging cloud platform, using a drone to photograph the area with haze in the current period, and obtaining a drone foggy image in the current period; Step S102: using drones to photograph historical areas with haze and historical areas without haze during the historical period, and collecting the photographed images to obtain historical drone photography records; Step S103: acquiring each historical drone shooting record stored in the cloud platform, and acquiring a historical drone foggy image from the historical drone shooting record; Step S104: analyzing the similarity between the drone foggy image and several historical drone foggy images in the historical drone shooting records in terms of the impact of fog on the image, wherein the specific process of analyzing the similarity between the drone foggy image and the a-th historical drone foggy image in the c-th historical drone shooting record in terms of the impact of fog on the image is to adjust the image size of the a-th historical drone foggy image based on the drone foggy image, perform image preprocessing on the a-th historical drone foggy image and the drone foggy image, and perform grayscale conversion on the drone foggy image and the a-th historical drone foggy image; Step S105: Obtain the dark channel image corresponding to the drone foggy image and the ath historical drone foggy image, wherein the dark channel image corresponding to the ath historical drone foggy image : ； in, An image window of pixel x in the a-th historical drone foggy image is preset; represents the gray value of pixel y in the a-th historical UAV foggy image; Step S106: Calculate the characteristic haze impact value of each pixel point of the a-th historical drone foggy image and the drone foggy image, wherein the characteristic haze impact value of the pixel point x in the a-th historical drone foggy image is : ； in, , ; is the preset characteristic haze constant; It is represented by the gray value of the pixel x in the a-th historical drone foggy image; Step S107: Calculate the characteristic haze impact similarity value between the drone foggy image and the ath historical drone foggy image : ； in, is the total number of pixels in the foggy image of the drone; is the characteristic haze impact value of the i-th pixel in the foggy image of the drone; is the characteristic haze impact value of the i-th pixel in the a-th historical UAV foggy image; Step S200: Based on the historical drone foggy image in the acquired historical drone shooting record, analyzing the image similarity between the historical drone foggy image and the drone foggy image to obtain a feature image similarity value; Step S300: Analyze the overall similarity between the historical drone foggy image and the drone foggy image based on the similarity value of the characteristic image haze effect between the historical drone foggy image and the drone foggy image, and obtain a characteristic historical drone foggy image in combination with the similarity degree value of the characteristic image between the historical drone foggy image and the drone foggy image; The step S300 includes: Step S301: Obtain the foggy image of the drone in the current period, and the haze effect similarity value of the characteristic image of the foggy image of several historical drones in the historical drone shooting records, and respectively obtain the maximum value of the haze effect similarity value of the foggy image of the drone and the characteristic image of several historical drones in the historical drone shooting records. and minimum value ; Step S302: Obtain the drone fog image in the current period, and the feature image similarity value of several historical drone fog images in each historical drone shooting record, and respectively obtain the maximum value of the feature image similarity value of the drone fog image and several historical drone fog images in each historical drone shooting record and minimum value ; Step S303: Analyze the overall similarity between the drone fog image and several historical drone fog images in each historical drone shooting record, wherein the drone fog image is similar to the first The first in the historic drone photography record The specific analysis process of the overall similarity between the historical drone foggy images is as follows: The first in the historic drone photography record The overall similarity of features between historical drone foggy images : ； in, is the foggy image of the drone, and The first in the historic drone photography record The similarity value of feature images between historical UAV foggy images; is the foggy image of the drone, and The first in the historic drone photography record The characteristic haze impact values ​​between the historical drone foggy images are similar; Step S304: When the drone has foggy image, The first in the historic drone photography record If the overall feature similarity value between the historical drone foggy images is greater than a preset overall feature similarity threshold, it is determined that the drone foggy image is similar to the first The first in the historic drone photography record The historical drone foggy images are similar in overall quality. historical drone foggy images, recorded as the characteristic historical drone foggy images of the drone foggy images; Step S400: obtaining a characteristic historical drone foggy image corresponding to the drone foggy image in the current period, obtaining a historical drone shooting record to which the characteristic historical drone foggy image belongs, optimizing the drone foggy image shot by the drone in the current period, and performing intelligent management on drone image defogging; The step S400 includes: Step S401: Acquire each characteristic historical drone foggy image of the drone foggy image in the current period, acquire the historical drone shooting record corresponding to each characteristic historical drone foggy image, and acquire the characteristic historical drone fog-free image of the historical area corresponding to the characteristic historical drone foggy image from the historical drone shooting record; Step S402: Based on the characteristic historical drone foggy images of the drone foggy image, and the characteristic historical drone fog-free images corresponding to the characteristic historical drone foggy images, the image of the drone foggy image in the current period after being defogged by the defogging algorithm is optimized, and various parameters of the algorithm are adjusted to perform intelligent management of drone image defogging.

2. The method for defogging drone images based on artificial intelligence according to claim 1, characterized in that: The step S200 includes: Step S201: acquiring each historical drone shooting record in the cloud platform, respectively acquiring a historical drone fog image of a historical area shot by a drone from each historical drone shooting record, and acquiring the drone fog image of the current period; Step S202: analyzing the image similarity between the drone foggy image and several historical drone foggy images in each historical drone shooting record, wherein the specific analysis process of the image similarity between the drone foggy image and the f-th historical drone foggy image in the e-th historical drone shooting record is to adjust the image size of the f-th historical drone foggy image according to the drone foggy image, perform image preprocessing on the adjusted f-th historical drone foggy image and the drone foggy image, and perform image grayscale conversion on the preprocessed f-th historical drone foggy image and the drone foggy image; Step S203: Calculate the brightness similarity value of the marked image between the drone foggy image and the f-th historical drone foggy image : ； in, is the average grayscale value of each pixel in the f-th historical UAV foggy image; is the average value of the grayscale values ​​of each pixel in the foggy image of the drone; is a preset first labeled image approximation constant; Step S204: Calculate the similarity value of the contrast between the labeled image of the drone in fog and the f-th historical drone in fog image : ； in, is the covariance between the drone foggy image and the f-th historical drone foggy image; is the variance of the f-th historical drone foggy image; is the variance of the foggy image of the drone; is a preset second marker image similarity constant; Step S205: Calculate the similarity value of the labeled image structure between the drone foggy image and the f-th historical drone foggy image : ； in, is the covariance of the f-th historical drone foggy image; is the covariance of the foggy image of the drone; is a preset third marker image similar constant; Step S206: Calculate the similarity value of the characteristic image between the drone foggy image and the f-th historical drone foggy image .

3. A drone image defogging system applied to the drone image defogging method based on artificial intelligence as described in any one of claims 1-2, characterized in that: The UAV image defogging system includes a haze impact degree analysis module, an image similarity degree analysis module, a feature history UAV foggy image module, and an image defogging intelligent management module; The haze impact degree analysis module is used to analyze the similarity between the historical drone foggy image and the drone foggy image in order to obtain a similar value of haze impact on the feature image; The image similarity analysis module is used to analyze the image similarity between the historical drone foggy image and the drone foggy image to obtain a feature image similarity value; The characteristic historical drone foggy image module is used to analyze the overall similarity between the historical drone foggy image and the drone foggy image to obtain the characteristic historical drone foggy image; The image defogging intelligent management module is used to obtain characteristic historical drone foggy images corresponding to drone foggy images in the current period, optimize the drone foggy images taken by drones in the current period, and perform intelligent management of drone image defogging.

4. The drone image defogging system according to claim 3, characterized in that: The haze impact degree analysis module includes a characteristic haze impact value unit and a characteristic haze impact similar value unit; The characteristic haze impact value unit is used to calculate the characteristic haze impact value of each pixel point of the historical drone foggy image and the drone foggy image; The characteristic haze impact similarity value unit is used to calculate the characteristic haze impact similarity value between the drone foggy image and the historical drone foggy images.

5. The drone image defogging system according to claim 3, characterized in that: The image similarity analysis module includes an image similarity analysis unit and a feature image similarity value unit; The image similarity analysis unit is used to analyze the image similarity between the drone foggy image and a plurality of historical drone foggy images in each historical drone shooting record; The characteristic image similarity value unit is used to calculate the characteristic image similarity value between the drone foggy image and the historical drone foggy images.

6. The drone image defogging system according to claim 3, characterized in that: The feature historical drone foggy image module includes a feature overall similarity value unit and a feature historical drone foggy image unit; The feature overall similarity value unit is used to calculate the feature overall similarity value between the drone foggy image and the historical drone foggy images in the historical drone shooting record; The characteristic historical drone foggy image unit is used to obtain the characteristic historical drone foggy image of the drone foggy image according to the characteristic overall similarity value between the drone foggy image and the historical drone foggy images in the historical drone shooting records.

7. The drone image defogging system according to claim 3, characterized in that: The image defogging intelligent management module includes an image defogging intelligent management unit; The image defogging intelligent management unit is used to optimize the image of the drone foggy image after defogging by the defogging algorithm in the current period according to each characteristic historical drone foggy image of the drone foggy image and the characteristic historical drone fog-free image corresponding to each characteristic historical drone foggy image, and adjust various parameters of the algorithm to perform intelligent management of drone image defogging.

Images