Image and video compression method and system based on drone front-end edge computing
By using image pixel value replacement mechanism and feature area merging technology in front-end computing of drone, the problem of low compression efficiency caused by the small proportion of similar images in the prior art is solved, and efficient compression and rapid transmission of image videos are achieved .
Patent Information
- Application Number
- CN202411155378.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-22
- Publication Date
- 2024-11-01
- Estimated Expiration
- 2044-08-22
AI Technical Summary
The existing image and video compression technology accounts for a small proportion of similar images in video or multi-frame images, resulting in a small compression amplitude and the overall file is still large, which cannot significantly improve the transmission speed.
Using the image video compression method based on the front-end edge calculation of the drone, the upper limit of the target pixel value difference of color recognition by the human eye is used to form an image pixel value replacement mechanism, and the feature points and feature areas are formed based on the minimum resolution area, and the The adjacent images are collected, the differential and common images are merged, the identification code is formed, and the boundary fit function is transmitted to the feature area.
By clustering and merging pixel points in the image, the amount of information is significantly reduced, the transmission efficiency is improved, and the needs of fast transmission are met.
Smart Images

Figure CN118678085B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of image processing technology, and in particular to an image and video compression method and system based on unmanned aerial vehicle front-end edge computing. Background Art
[0002] The images obtained by the drone are videos or pictures formed by the continuous action of several frames of static pictures. As the image quality resolution increases, the volume occupied by a single frame of static pictures also increases. In addition, a video contains a large number of static pictures, which will cause the file size to be large. A file that is too large will not only take up too much hard disk space, but also be inconvenient for fast network transmission, resulting in the recipient being unable to quickly receive the file and analyze it.
[0003] Existing compression technology usually uses similar frame image recognition methods to deduplicate and compress similar images. However, in videos or multi-frame images, the proportion of similar images is small. Therefore, the compression range is small, the overall file is still large, and the transmission speed cannot be significantly improved. Summary of the invention
[0004] In order to solve the above technical problems, an image and video compression method and system based on drone front-end edge computing are provided. The technical solution solves the problem that the existing compression technology proposed in the above background technology usually adopts the method of similar frame image recognition to deduplicate and compress similar images, but in videos or multi-frame images, the proportion of similar images is small. Therefore, the compression range is small, the overall file is still large, and the transmission speed cannot be significantly improved.
[0005] In order to achieve the above purpose, the technical solution adopted by the present invention is:
[0006] An image and video compression method based on unmanned aerial vehicle front-end edge computing, comprising:
[0007] Acquire at least one image to be compressed;
[0008] Determine the upper limit of the target pixel value difference for color recognition by the human eye, and form an image pixel value replacement mechanism based on the upper limit of the target pixel value difference;
[0009] Determine the minimum resolution area of the image for the human eye;
[0010] Using the image pixel replacement mechanism, the pixel values of the pixels in the image to be compressed are replaced to obtain a preprocessed image;
[0011] In the preprocessed image, at least one feature point and a feature area are formed based on the minimum resolution area;
[0012] Dividing the preprocessed image into at least one adjacent image set in arrangement order;
[0013] Obtaining the overlapping parts of the pre-processed images in the adjacent image set as a common image, and taking the parts of the pre-processed images other than the common image as a difference image, and forming an identification code for the common image;
[0014] Merging at least one feature point and feature region of the difference image and at least one feature point and feature region of the common image to obtain a target feature point and a target feature region;
[0015] A boundary fitting function is formed for the target feature area, the coordinates and pixel values of the target feature points are encoded and transmitted, and the boundary fitting function of the target feature area is encoded and transmitted;
[0016] A video to be compressed is obtained, the video to be compressed is divided into frames to obtain at least one video frame image, and the at least one video frame image is compressed using the step of compressing at least one image to be compressed.
[0017] Preferably, determining the upper limit of the target pixel value difference for color recognition by the human eye comprises the following steps:
[0018] Acquire a sample pixel point, generate at least one simulated pixel point based on the pixel value of the sample pixel point, and arrange the simulated pixel points in ascending order according to the pixel value of the simulated pixel point, wherein the difference between the pixel values of adjacent simulated pixel points is a preset pixel value;
[0019] Obtain at least one unrecognizable simulated pixel point that cannot be distinguished by human eyes from among the sample pixels;
[0020] Obtaining an unrecognizable simulated pixel point with the largest pixel value among at least one unrecognizable simulated pixel point as a target unrecognizable simulated pixel point;
[0021] The difference between the pixel value of the target unrecognizable simulated pixel point and the pixel value of the sample pixel point is used as the upper limit of the target pixel value difference;
[0022] The steps of obtaining at least one unrecognizable simulated pixel point that cannot be distinguished by human eyes among the sample pixels are as follows:
[0023] The simulated pixel is mixed into at least one confused pixel, and the confused pixel is generated by copying the sample pixel. When the human eye cannot identify the simulated pixel from the simulated pixel and the confused pixel, the simulated pixel is regarded as an unrecognizable simulated pixel, otherwise, no processing is performed.
[0024] Preferably, the image pixel value replacement mechanism formed based on the target pixel value difference upper limit comprises the following steps:
[0025] When the distance between the two pixel points is less than the preset distance, if the difference in pixel values of the two pixel points is less than the target pixel value difference upper limit, any one of the two pixel points is used to replace the two pixel points.
[0026] Preferably, determining the minimum resolution area of the image by the human eye comprises the following steps:
[0027] Generate at least one simulation area, and arrange the simulation areas in ascending order according to the area of the simulation areas, wherein the simulation areas are squares, and the difference in the side lengths of adjacent simulation areas is a preset length;
[0028] Assign sample pixel values to all pixel points in the simulation area, and obtain a simulation area recognizable by human eyes as the target simulation area;
[0029] The target simulation area with the smallest area is taken as the minimum resolution area;
[0030] The steps of obtaining the simulated area recognizable by the human eye are as follows:
[0031] The simulation area is mixed into at least one confusing area. When the pixel points of the simulation area can be correctly identified in at least one confusing area, the simulation area is used as the target simulation area, otherwise, no processing is performed.
[0032] Preferably, replacing the pixel values of the pixels in the image to be compressed to obtain the preprocessed image comprises the following steps:
[0033] Segmenting the image to be compressed with a preset distance as the length to obtain at least one pre-processed region, wherein the side lengths of the pre-processed regions are all the preset distances;
[0034] Classifying the pixels in the preprocessing area to obtain at least one pixel point set, wherein the pixel value difference of the pixels in the pixel point set is less than the target pixel value difference upper limit;
[0035] Use the image pixel value replacement mechanism to replace the pixels in the pixel set with the same pixel.
[0036] Preferably, the forming of at least one feature point and feature area based on the minimum resolution area comprises the following steps:
[0037] Segmenting the preprocessed image according to the shape of the minimum resolution area to obtain at least one area to be processed;
[0038] Count the pixel values of the pixels in the area to be processed, and obtain the pixel value with the largest proportion as the calibration pixel value;
[0039] The pixel values of the pixels in the area to be processed are all adjusted to the calibration pixel values to obtain the feature area;
[0040] The center of the feature area is taken as the feature point, and coordinate modeling is performed on all preprocessed images in the same way to obtain the coordinates of the feature point.
[0041] Preferably, dividing the pre-processed image into at least one adjacent image set in the order of arrangement comprises the following steps:
[0042] Arrange at least one preprocessed image in sequence, and obtain at least one positioning preprocessed image from at least one preprocessed image, wherein the similarity between the preprocessed image between adjacent positioning preprocessed images and the positioning preprocessed image on the left side of the adjacent positioning preprocessed images is less than a preset value, and the similarity between the adjacent positioning preprocessed images is greater than a preset value;
[0043] Aggregating the preprocessed images between the adjacent positioning preprocessed images and the positioning preprocessed image on the left side of the adjacent positioning preprocessed images to obtain a set of adjacent images;
[0044] The similarity of the images is calculated as follows: the absolute values of the differences between the pixel values of the corresponding pixels in the two images are accumulated to obtain the similarity of the images.
[0045] Preferably, the step of merging at least one feature point and feature region of the difference image and at least one feature point and feature region of the common image to obtain a target feature point and target feature region comprises the following steps:
[0046] In the difference image, when two feature regions are adjacent and pixel values of pixel points of the two feature regions are consistent, the two feature regions are merged;
[0047] When the feature regions in the difference image are merged, the merged feature region is used as the target feature region, and the center of the merged feature region is used as the target feature point;
[0048] In the common image, when two feature regions are adjacent and pixel values of pixel points of the two feature regions are consistent, the two feature regions are merged;
[0049] When the merging of the feature regions in the common image is completed, the merged feature region is used as the target feature region, and the center of the merged feature region is used as the target feature point.
[0050] Preferably, forming a boundary fitting function for the target feature area comprises the following steps:
[0051] At least one sampling point is taken at the edge of the target feature area, the coordinates of the sampling point are obtained, and the sampling point is fitted to obtain a boundary fitting function.
[0052] An image and video compression system based on drone front-end edge computing is used to implement the above-mentioned image and video compression method based on drone front-end edge computing, including:
[0053] An image acquisition module, wherein the image acquisition module acquires at least one image to be compressed;
[0054] A data acquisition module, wherein the data acquisition module determines an upper limit of a target pixel value difference for color recognition by the human eye, forms an image pixel value replacement mechanism based on the upper limit of the target pixel value difference, and determines a minimum resolution area of the image for the human eye;
[0055] A preprocessing module, wherein the preprocessing module uses an image pixel replacement mechanism to replace pixel values of pixels in the image to be compressed to obtain a preprocessed image;
[0056] A feature forming module, wherein the feature forming module forms at least one feature point and a feature area in the preprocessed image based on the minimum resolution area;
[0057] An image division module, wherein the image division module divides the preprocessed image into at least one adjacent image set according to an arrangement order;
[0058] An image comparison module, wherein the image comparison module obtains an overlapping portion of preprocessed images in a set of adjacent images as a common image, and uses a portion of the preprocessed images other than the common image as a difference image to form an identification code for the common image;
[0059] A region merging module, which merges at least one feature point and feature region of the difference image and at least one feature point and feature region in the common image to obtain a target feature point and a target feature region;
[0060] An image transmission module, wherein the image transmission module forms a boundary fitting function for the target feature area, encodes the coordinates and pixel values of the target feature points and transmits them, and encodes the boundary fitting function for the target feature area and transmits them;
[0061] A video compression module obtains a video to be compressed, divides the video to be compressed into frames, obtains at least one video frame image, and compresses at least one video frame image using the step of compressing at least one image to be compressed.
[0062] Compared with the prior art, the present invention has the following beneficial effects:
[0063] By setting up a data acquisition module, a preprocessing module, a feature formation module and a region merging module, the physical limitations of the human eye on image recognition are used to homogenize the differences that cannot be recognized by the human eye. The pixels in the image can be clustered and represented by the pixel value of the same pixel. Since each region is represented by only a single pixel, the amount of information that needs to be transmitted in each image can be greatly reduced. When the common image and the difference image are obtained, the feature areas in the common image and the difference image are merged. This can further reduce the amount of information that needs to be transmitted. At the same time, the image can be restored based on the transmitted information, thereby meeting the transmission requirements and significantly improving the transmission efficiency. BRIEF DESCRIPTION OF THE DRAWINGS
[0064] Figure 1 It is a flow chart of the image and video compression method based on the front-end edge computing of the unmanned aerial vehicle of the present invention;
[0065] Figure 2 A schematic diagram of a flow chart of determining the upper limit of the target pixel value difference for color recognition by the human eye according to the present invention;
[0066] Figure 3 A schematic diagram of a process of determining the minimum resolution area of an image by the human eye according to the present invention;
[0067] Figure 4 A schematic diagram of a process of replacing pixel values of pixels in an image to be compressed to obtain a preprocessed image according to the present invention;
[0068] Figure 5 A schematic diagram of a process of forming at least one feature point and a feature area based on a minimum resolution area of the present invention;
[0069] Figure 6 A schematic diagram of a process of dividing a pre-processed image into at least one adjacent image set according to an arrangement order of the present invention;
[0070] Figure 7 The present invention is a flow chart of merging at least one feature point and feature region of a difference image and at least one feature point and feature region of a common image to obtain a target feature point and a target feature region. DETAILED DESCRIPTION
[0071] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art may think of other obvious variations.
[0072] Reference Figure 1 As shown, an image and video compression method based on drone front-end edge computing includes:
[0073] Acquire at least one image to be compressed;
[0074] Determine the upper limit of the target pixel value difference for color recognition by the human eye, and form an image pixel value replacement mechanism based on the upper limit of the target pixel value difference;
[0075] Determine the minimum resolution area of the image for the human eye;
[0076] Using the image pixel replacement mechanism, the pixel values of the pixels in the image to be compressed are replaced to obtain a preprocessed image;
[0077] In the preprocessed image, at least one feature point and a feature area are formed based on the minimum resolution area;
[0078] Dividing the preprocessed image into at least one adjacent image set in arrangement order;
[0079] Obtaining the overlapping parts of the pre-processed images in the adjacent image set as a common image, and taking the parts of the pre-processed images other than the common image as a difference image, and forming an identification code for the common image;
[0080] Merging at least one feature point and feature region of the difference image and at least one feature point and feature region of the common image to obtain a target feature point and a target feature region;
[0081] A boundary fitting function is formed for the target feature area, the coordinates and pixel values of the target feature points are encoded and transmitted, and the boundary fitting function of the target feature area is encoded and transmitted;
[0082] A video to be compressed is obtained, the video to be compressed is divided into frames to obtain at least one video frame image, and the at least one video frame image is compressed using the step of compressing at least one image to be compressed.
[0083] When obtaining the overlapping part of the preprocessed images in the adjacent image set, the pixel points at the corresponding positions in the preprocessed images in the adjacent image set are compared. When the pixel values of the pixel points at the corresponding positions are consistent, the pixel points are regarded as the overlapping part of the preprocessed images in the adjacent image set, and then a common image is obtained, and the part of the preprocessed images in the adjacent image set outside the common image is the difference image;
[0084] The common image of the pre-processed images in the adjacent image set only needs to be transmitted once, which can reduce the amount of transmitted data, while the difference images need to be transmitted separately.
[0085] Reference Figure 2 As shown, determining the upper limit of the target pixel value difference for color recognition by the human eye includes the following steps:
[0086] Acquire a sample pixel point, generate at least one simulated pixel point based on the pixel value of the sample pixel point, and arrange the simulated pixel points in ascending order according to the pixel value of the simulated pixel point, wherein the difference between the pixel values of adjacent simulated pixel points is a preset pixel value;
[0087] Obtain at least one unrecognizable simulated pixel point that cannot be distinguished by human eyes from among the sample pixels;
[0088] Obtaining an unrecognizable simulated pixel point with the largest pixel value among at least one unrecognizable simulated pixel point as a target unrecognizable simulated pixel point;
[0089] The difference between the pixel value of the target unrecognizable simulated pixel point and the pixel value of the sample pixel point is used as the upper limit of the target pixel value difference;
[0090] The steps of obtaining at least one unrecognizable simulated pixel point that cannot be distinguished by human eyes among the sample pixels are as follows:
[0091] The simulated pixel is mixed into at least one confused pixel, and the confused pixel is generated by copying the sample pixel. When the human eye cannot identify the simulated pixel from the simulated pixel and the confused pixel, the simulated pixel is regarded as an unrecognizable simulated pixel, otherwise, no processing is performed.
[0092] The human eye has a physical limit in recognizing the difference between pixels. When the difference in pixel values is small enough, the difference between the two pixels cannot be recognized by the human eye. Therefore, when performing difference processing, the amount of data transmitted will increase. Therefore, in order to reduce the amount of data transmitted, a replacement mechanism is formed to replace pixels with small differences with the same pixel value. At the same time, it will not affect the restoration of the overall image.
[0093] Based on the target pixel value difference upper limit, the image pixel value replacement mechanism is formed, including the following steps:
[0094] When the distance between the two pixel points is less than the preset distance, if the difference in pixel values of the two pixel points is less than the target pixel value difference upper limit, any one of the two pixel points is used to replace the two pixel points.
[0095] Reference Figure 3 As shown, determining the minimum resolution area of the image by the human eye includes the following steps:
[0096] Generate at least one simulation area, and arrange the simulation areas in ascending order according to the area of the simulation areas, wherein the simulation areas are squares, and the difference in the side lengths of adjacent simulation areas is a preset length;
[0097] Assign sample pixel values to all pixel points in the simulation area, and obtain a simulation area recognizable by human eyes as the target simulation area;
[0098] The target simulation area with the smallest area is taken as the minimum resolution area;
[0099] The steps of obtaining the simulated area recognizable by the human eye are as follows:
[0100] The simulation area is mixed into at least one confusing area. When the pixel points of the simulation area can be correctly identified in at least one confusing area, the simulation area is used as the target simulation area, otherwise, no processing is performed.
[0101] The recognition area of the human eye has physical limits. When the area is extremely small, the pixels in the area cannot be recognized. Therefore, covering the area with the pixels that account for the largest proportion of the area will not affect the recognition of the human eye. However, since the pixels in the area are unified, only one pixel and the position of the area need to be recorded for the area, which can further reduce the amount of data that needs to be transmitted.
[0102] Reference Figure 4 As shown, replacing the pixel values of the pixels in the compressed image to obtain the preprocessed image includes the following steps:
[0103] Segmenting the image to be compressed with a preset distance as the length to obtain at least one pre-processed region, wherein the side lengths of the pre-processed regions are all the preset distances;
[0104] Classifying the pixels in the preprocessing area to obtain at least one pixel point set, wherein the pixel value difference of the pixels in the pixel point set is less than the target pixel value difference upper limit;
[0105] Use the image pixel value replacement mechanism to replace the pixels in the pixel set with the same pixel.
[0106] Reference Figure 5 As shown, based on the minimum resolution area, forming at least one feature point and a feature area includes the following steps:
[0107] Segmenting the preprocessed image according to the shape of the minimum resolution area to obtain at least one area to be processed;
[0108] Count the pixel values of the pixels in the area to be processed, and obtain the pixel value with the largest proportion as the calibration pixel value;
[0109] The pixel values of the pixels in the area to be processed are all adjusted to the calibration pixel values to obtain the feature area;
[0110] The center of the feature area is taken as the feature point, and coordinate modeling is performed on all preprocessed images in the same way to obtain the coordinates of the feature point.
[0111] Reference Figure 6As shown, dividing the pre-processed image into at least one adjacent image set in the order of arrangement includes the following steps:
[0112] Arrange at least one preprocessed image in sequence, and obtain at least one positioning preprocessed image from at least one preprocessed image, wherein the similarity between the preprocessed image between adjacent positioning preprocessed images and the positioning preprocessed image on the left side of the adjacent positioning preprocessed images is less than a preset value, and the similarity between the adjacent positioning preprocessed images is greater than a preset value;
[0113] Aggregating the preprocessed images between the adjacent positioning preprocessed images and the positioning preprocessed image on the left side of the adjacent positioning preprocessed images to obtain a set of adjacent images;
[0114] The similarity of the images is calculated as follows: the absolute values of the differences between the pixel values of the corresponding pixels in the two images are accumulated to obtain the similarity of the images.
[0115] The classification purpose is: in the order of pre-processed images, at least one continuous pre-processed image adjacent to the positioning pre-processed image is obtained, thereby forming a set of adjacent images. Since the order of formation is continuous, it is easier to restore after transmission;
[0116] Here, the left side means that the images are arranged in the order of preprocessed images and are placed on the left side of the sequence.
[0117] Reference Figure 7 As shown, merging at least one feature point and feature region of the difference image and at least one feature point and feature region in the common image to obtain a target feature point and target feature region includes the following steps:
[0118] In the difference image, when two feature regions are adjacent and pixel values of pixel points of the two feature regions are consistent, the two feature regions are merged;
[0119] When the feature regions in the difference image are merged, the merged feature region is used as the target feature region, and the center of the merged feature region is used as the target feature point;
[0120] In the common image, when two feature regions are adjacent and pixel values of pixel points of the two feature regions are consistent, the two feature regions are merged;
[0121] When the merging of the feature regions in the common image is completed, the merged feature region is used as the target feature region, and the center of the merged feature region is used as the target feature point.
[0122] Due to the limitation of feature region division, some adjacent feature regions are not merged even though the pixels inside them have the same pixel value. Therefore, merging them into the entire region can further reduce the number of feature regions and thus reduce the amount of data transmitted.
[0123] Forming a boundary fitting function for the target feature area includes the following steps:
[0124] At least one sampling point is taken at the edge of the target feature area, the coordinates of the sampling point are obtained, and the sampling point is fitted to obtain a boundary fitting function.
[0125] The purpose of the boundary fitting function is to restore the target feature area. Otherwise, since the target feature area is the area after the feature areas are merged, its shape is uncertain and cannot be directly restored based on the target feature points.
[0126] An image and video compression system based on drone front-end edge computing is used to implement the above-mentioned image and video compression method based on drone front-end edge computing, including:
[0127] An image acquisition module, wherein the image acquisition module acquires at least one image to be compressed;
[0128] A data acquisition module, wherein the data acquisition module determines an upper limit of a target pixel value difference for color recognition by the human eye, forms an image pixel value replacement mechanism based on the upper limit of the target pixel value difference, and determines a minimum resolution area of the image for the human eye;
[0129] A preprocessing module, wherein the preprocessing module uses an image pixel replacement mechanism to replace pixel values of pixels in the image to be compressed to obtain a preprocessed image;
[0130] A feature forming module, wherein the feature forming module forms at least one feature point and a feature area in the preprocessed image based on the minimum resolution area;
[0131] An image division module, wherein the image division module divides the preprocessed image into at least one adjacent image set according to an arrangement order;
[0132] An image comparison module, wherein the image comparison module obtains an overlapping portion of preprocessed images in a set of adjacent images as a common image, and uses a portion of the preprocessed images other than the common image as a difference image to form an identification code for the common image;
[0133] A region merging module, which merges at least one feature point and feature region of the difference image and at least one feature point and feature region in the common image to obtain a target feature point and a target feature region;
[0134] An image transmission module, wherein the image transmission module forms a boundary fitting function for the target feature area, encodes the coordinates and pixel values of the target feature points and transmits them, and encodes the boundary fitting function for the target feature area and transmits them;
[0135] A video compression module obtains a video to be compressed, divides the video to be compressed into frames, obtains at least one video frame image, and compresses at least one video frame image using the step of compressing at least one image to be compressed.
[0136] Furthermore, the present solution also proposes a storage medium on which a computer-readable program is stored. When the computer-readable program is called, the above-mentioned image and video compression method based on drone front-end edge computing is executed.
[0137] It is understandable that the storage medium may be a magnetic medium, such as a floppy disk, a hard disk, or a magnetic tape; an optical medium, such as a DVD; or a semiconductor medium, such as a solid state drive (SSD).
[0138] In summary, the advantages of the present invention are: by setting a data acquisition module, a preprocessing module, a feature formation module and a region merging module, the differences that cannot be recognized by the human eye are homogenized through the physical limitations of the human eye on image recognition, so that the pixels in the image can be clustered and represented by the pixel value of the same pixel. Since each area is represented by only a single pixel, the amount of information that needs to be transmitted in each image can be greatly reduced. When the common image and the difference image are obtained, the feature areas in the common image and the difference image are merged, thereby further reducing the amount of information that needs to be transmitted. At the same time, the image can be restored according to the transmitted information, thereby meeting the transmission requirements and significantly improving the transmission efficiency.
[0139] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions only describe the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention. These changes and improvements fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the attached claims and their equivalents.
Claims
1. An image and video compression method based on drone front-end edge computing, characterized in that: include: Acquire at least one image to be compressed; Determine the upper limit of the target pixel value difference for color recognition by the human eye, and form an image pixel value replacement mechanism based on the upper limit of the target pixel value difference; Determine the minimum resolution area of the image for the human eye; Using the image pixel replacement mechanism, the pixel values of the pixels in the image to be compressed are replaced to obtain a preprocessed image; In the preprocessed image, at least one feature point and a feature area are formed based on the minimum resolution area; Dividing the preprocessed image into at least one adjacent image set according to the arrangement order; Obtaining the overlapping parts of the pre-processed images in the adjacent image set as a common image, and taking the parts of the pre-processed images other than the common image as a difference image, and forming an identification code for the common image; Merging at least one feature point and feature region of the difference image and at least one feature point and feature region of the common image to obtain a target feature point and a target feature region; A boundary fitting function is formed for the target feature area, the coordinates and pixel values of the target feature points are encoded and transmitted, and the boundary fitting function of the target feature area is encoded and transmitted; Acquire a video to be compressed, divide the video to be compressed into frames to obtain at least one video frame image, and compress the at least one video frame image using the step of compressing at least one image to be compressed; Determining the upper limit of the target pixel value difference for color recognition by the human eye comprises the following steps: Acquire a sample pixel point, generate at least one simulated pixel point based on the pixel value of the sample pixel point, and arrange the simulated pixel points in ascending order according to the pixel value of the simulated pixel point, wherein the difference between the pixel values of adjacent simulated pixel points is a preset pixel value; Obtain at least one unrecognizable simulated pixel point that cannot be distinguished by human eyes from among the sample pixels; Obtaining an unrecognizable simulated pixel point with the largest pixel value among at least one unrecognizable simulated pixel point as a target unrecognizable simulated pixel point; The difference between the pixel value of the target unrecognizable simulated pixel point and the pixel value of the sample pixel point is used as the upper limit of the target pixel value difference; The steps of obtaining at least one unrecognizable simulated pixel point that cannot be distinguished by human eyes among the sample pixels are as follows: Mixing the simulated pixel into at least one confused pixel, where the confused pixel is generated by copying the sample pixel. When the human eye cannot identify the simulated pixel from the simulated pixel and the confused pixel, the simulated pixel is regarded as an unrecognizable simulated pixel. Otherwise, no processing is performed. The replacement mechanism for forming the image pixel value based on the target pixel value difference upper limit comprises the following steps: When the distance between the two pixel points is less than the preset distance, if the difference between the pixel values of the two pixel points is less than the target pixel value difference upper limit, use any one of the two pixel points to replace the two pixel points; Determining the minimum resolution area of the image by the human eye comprises the following steps: Generate at least one simulation area, and arrange the simulation areas in ascending order according to the area of the simulation areas, wherein the simulation areas are squares, and the difference in the side lengths of adjacent simulation areas is a preset length; Assign sample pixel values to all pixel points in the simulation area, and obtain a simulation area recognizable by human eyes as the target simulation area; The target simulation area with the smallest area is taken as the minimum resolution area; The steps of obtaining the simulated area recognizable by the human eye are as follows: The simulation area is mixed into at least one confusion area. When the pixel points of the simulation area can be correctly identified in at least one confusion area, the simulation area is used as the target simulation area, otherwise, no processing is performed; The step of replacing the pixel values of the pixels in the image to be compressed to obtain the preprocessed image comprises the following steps: Segmenting the image to be compressed with a preset distance as the length to obtain at least one pre-processed region, wherein the side lengths of the pre-processed regions are all the preset distances; Classifying the pixels in the preprocessing area to obtain at least one pixel point set, wherein the pixel value difference of the pixels in the pixel point set is less than the target pixel value difference upper limit; Use the image pixel value replacement mechanism to replace the pixels in the pixel set with the same pixel; The forming of at least one feature point and a feature area based on the minimum resolution area comprises the following steps: Segmenting the preprocessed image according to the shape of the minimum resolution area to obtain at least one area to be processed; Count the pixel values of the pixels in the area to be processed, and obtain the pixel value with the largest proportion as the calibration pixel value; The pixel values of the pixels in the area to be processed are all adjusted to the calibration pixel values to obtain the feature area; The center of the feature area is taken as the feature point, and coordinate modeling is performed on all preprocessed images in the same way to obtain the coordinates of the feature point.
2. The image and video compression method based on drone front-end edge computing according to claim 1 is characterized in that: The step of merging at least one feature point and feature region in the difference image and at least one feature point and feature region in the common image to obtain a target feature point and target feature region comprises the following steps: In the difference image, when two feature regions are adjacent and pixel values of pixel points of the two feature regions are consistent, the two feature regions are merged; When the feature regions in the difference image are merged, the merged feature region is used as the target feature region, and the center of the merged feature region is used as the target feature point; In the common image, when two feature regions are adjacent and pixel values of pixel points of the two feature regions are consistent, the two feature regions are merged; When the merging of the feature regions in the common image is completed, the merged feature region is used as the target feature region, and the center of the merged feature region is used as the target feature point.
3. The image and video compression method based on drone front-end edge computing according to claim 2 is characterized in that: The forming of the boundary fitting function for the target feature region comprises the following steps: At least one sampling point is taken at the edge of the target feature area, the coordinates of the sampling point are obtained, and the sampling point is fitted to obtain a boundary fitting function.
4. An image and video compression system based on drone front-end edge computing, used to implement the image and video compression method based on drone front-end edge computing as described in any one of claims 1 to 3, characterized in that: include: An image acquisition module, wherein the image acquisition module acquires at least one image to be compressed; A data acquisition module, wherein the data acquisition module determines an upper limit of a target pixel value difference for color recognition by the human eye, forms an image pixel value replacement mechanism based on the upper limit of the target pixel value difference, and determines a minimum resolution area of the image for the human eye; A preprocessing module, wherein the preprocessing module uses an image pixel replacement mechanism to replace pixel values of pixels in the image to be compressed to obtain a preprocessed image; A feature forming module, wherein the feature forming module forms at least one feature point and a feature area in the preprocessed image based on the minimum resolution area; An image division module, wherein the image division module divides the preprocessed image into at least one adjacent image set according to an arrangement order; An image comparison module, wherein the image comparison module obtains an overlapping portion of preprocessed images in a set of adjacent images as a common image, and uses a portion of the preprocessed images other than the common image as a difference image to form an identification code for the common image; A region merging module, which merges at least one feature point and feature region of the difference image and at least one feature point and feature region in the common image to obtain a target feature point and a target feature region; An image transmission module, wherein the image transmission module forms a boundary fitting function for the target feature area, encodes the coordinates and pixel values of the target feature points and transmits them, and encodes the boundary fitting function for the target feature area and transmits them; A video compression module, wherein the video compression module obtains a video to be compressed, divides the video to be compressed into frames, obtains at least one video frame image, and compresses the at least one video frame image using the step of compressing at least one image to be compressed; Determining the upper limit of the target pixel value difference for color recognition by the human eye comprises the following steps: Acquire a sample pixel point, generate at least one simulated pixel point based on the pixel value of the sample pixel point, and arrange the simulated pixel points in ascending order according to the pixel value of the simulated pixel point, wherein the difference between the pixel values of adjacent simulated pixel points is a preset pixel value; Obtain at least one unrecognizable simulated pixel point that cannot be distinguished by human eyes from among the sample pixels; Obtaining an unrecognizable simulated pixel point with the largest pixel value among at least one unrecognizable simulated pixel point as a target unrecognizable simulated pixel point; The difference between the pixel value of the target unrecognizable simulated pixel point and the pixel value of the sample pixel point is used as the upper limit of the target pixel value difference; The steps of obtaining at least one unrecognizable simulated pixel point that cannot be distinguished by human eyes among the sample pixels are as follows: Mixing the simulated pixel into at least one confused pixel, where the confused pixel is generated by copying the sample pixel. When the human eye cannot identify the simulated pixel from the simulated pixel and the confused pixel, the simulated pixel is regarded as an unrecognizable simulated pixel. Otherwise, no processing is performed. The replacement mechanism for forming the image pixel value based on the target pixel value difference upper limit comprises the following steps: When the distance between the two pixel points is less than the preset distance, if the difference between the pixel values of the two pixel points is less than the target pixel value difference upper limit, use any one of the two pixel points to replace the two pixel points; Determining the minimum resolution area of the image by the human eye comprises the following steps: Generate at least one simulation area, and arrange the simulation areas in ascending order according to the area of the simulation areas, wherein the simulation areas are squares, and the difference in the side lengths of adjacent simulation areas is a preset length; Assign sample pixel values to all pixel points in the simulation area, and obtain a simulation area recognizable by human eyes as the target simulation area; The target simulation area with the smallest area is taken as the minimum resolution area; The steps of obtaining the simulated area recognizable by the human eye are as follows: The simulation area is mixed into at least one confusion area. When the pixel points of the simulation area can be correctly identified in at least one confusion area, the simulation area is used as the target simulation area, otherwise, no processing is performed; The step of replacing the pixel values of the pixels in the image to be compressed to obtain the preprocessed image comprises the following steps: Segmenting the image to be compressed with a preset distance as the length to obtain at least one pre-processed region, wherein the side lengths of the pre-processed regions are all the preset distances; Classifying the pixels in the preprocessing area to obtain at least one pixel point set, wherein the pixel value difference of the pixels in the pixel point set is less than the target pixel value difference upper limit; Use the image pixel value replacement mechanism to replace the pixels in the pixel set with the same pixel; The forming of at least one feature point and a feature area based on the minimum resolution area comprises the following steps: Segmenting the preprocessed image according to the shape of the minimum resolution area to obtain at least one area to be processed; Count the pixel values of the pixels in the area to be processed, and obtain the pixel value with the largest proportion as the calibration pixel value; The pixel values of the pixels in the area to be processed are all adjusted to the calibration pixel values to obtain the feature area; The center of the feature area is taken as the feature point, and coordinate modeling is performed on all preprocessed images in the same way to obtain the coordinates of the feature point.
