Edge gateway-based satellite image transmission video processing method and device and medium

Abstract

The application discloses a satellite image transmission video processing method and device based on an edge gateway and a medium. The method comprises the following steps: acquiring a video frame sequence of local monitoring through a sending end, performing content motion feature analysis on the video frame sequence based on a motion analysis model to determine a content motion feature gear, then performing frame extraction and image preprocessing on the video frame sequence according to the content motion feature gear, encoding the video frame sequence to generate a compressed video code stream, packaging the compressed video code stream, encoding parameters and the number of frames into a compressed video code stream container, and uploading the compressed video code stream through a satellite link. The receiving end receives the compressed video code stream, analyzes, decodes and post-processes the compressed video code stream, enlarges the compressed video code stream to a display resolution, and then outputs the compressed video code stream as a playable video file. The application can solve the problems that the fixed encoding code rate of existing video data cannot dynamically match the motion features of monitoring video during satellite image transmission, and the decoding strategy of the receiving end cannot be flexibly configured.

Description

Technical Field

[0001] This invention relates to the field of satellite image transmission data processing technology, and in particular to satellite image transmission video processing methods, apparatus and media based on edge gateways. Background Technology

[0002] In a satellite image transmission system, the transmitting end, deployed at an edge gateway, is responsible for collecting or receiving surveillance video collected by data acquisition terminals, compressing the surveillance video, uplinking it through the satellite link, and then receiving the compressed surveillance video at the receiving end to decompress it and restore the surveillance video, thereby providing data support for subsequent data processing. However, when edge gateways compress video data, they typically use a fixed constant bitrate or constant quality factor to standardly encode the original resolution surveillance video. This encoding method has several drawbacks. When a low bitrate is retained during encoding, the encoder is forced to perform extensive quantization, resulting in significant block artifacts and loss of detail in the compressed video data. Conversely, when a high bitrate is retained, the encoded data size cannot meet the size constraints of satellite short messages or packetized links, leading to incomplete data. Furthermore, using a constant bitrate for video data encoding does not differentiate between static and high-dynamic scenes, resulting in too many redundant frames in static scenes and potentially undecoding artifacts in dynamic scenes due to insufficient bitrate. Moreover, since the standard decoder at the transmitting end outputs distorted encoded resolution images, the receiving end must strictly employ the same decoding strategy as the transmitting end for post-processing to ensure a clear image and minimal block artifacts for the user. This strict requirement for the receiving end's decoding strategy to match the transmitting end hinders flexible configuration at the receiving end. Summary of the Invention

[0003] To overcome the shortcomings of existing technologies, one of the objectives of this invention is to provide a satellite image transmission video processing method based on an edge gateway. This method can solve the problems that existing video uses a fixed encoding bitrate during satellite image transmission, resulting in a mismatch between the encoding volume and the volume constraints of satellite image transmission. It also addresses the issues that the encoding bitrate cannot be dynamically adapted to the dynamic and static types of the monitoring video, and that the strict matching between the encoding strategy of the transmitting end and the decoding strategy of the receiving end is not conducive to the flexible configuration of the receiving end.

[0004] The second objective of this invention is to provide a satellite image transmission video processing device based on an edge gateway, which can solve the problems of existing video using a fixed encoding bitrate during satellite image transmission, resulting in a mismatch between the encoding volume and the volume constraints of satellite image transmission, the inability of the encoding bitrate to dynamically adapt to the dynamic and static types of the monitoring video, and the strict matching between the encoding strategy of the transmitting end and the decoding strategy of the receiving end, which is not conducive to the flexible configuration of the receiving end.

[0005] The third objective of this invention is to provide a computer-readable storage medium that can solve the problems of existing video transmission using a fixed encoding bitrate, which leads to a mismatch between the encoded size and the size constraint of satellite transmission, the inability of the encoding bitrate to dynamically adapt to the static and dynamic types of the surveillance video, and the strict matching between the encoding strategy of the transmitting end and the decoding strategy of the receiving end, which is not conducive to the flexible configuration of the receiving end.

[0006] One of the objectives of this invention is achieved through the following technical solution: A satellite image transmission video processing method based on an edge gateway, the satellite image transmission video processing method comprising: Video frame acquisition steps: Acquire the video frame sequence of local monitoring through the transmitter deployed on the first edge gateway, the video frame sequence including multiple consecutive video frames; Adaptive frame extraction step: The transmitting end performs content motion feature analysis on the video frame sequence based on a preset motion analysis model to determine the content motion feature level, and then extracts frames from the video frame sequence according to the content motion feature level to obtain a sparse frame sequence. Preprocessing steps: The sparse frame sequence is geometrically downsampled by the transmitting end and converted into a frame sequence at the coding resolution, and the frame sequence at the coding resolution is deblocked. Encoding steps: The transmitting end determines the variable bitrate encoding of the standard video file according to the preset target volume and content motion feature level. Then, the deblocked frame sequence is encoded according to the variable bitrate encoding of the standard video file to generate a compressed video stream. The compressed video stream, encoding parameters and frame number are encapsulated into a compressed video stream container, and the compressed video stream container is uplinked through the satellite link. Decoding steps: The compressed video stream container is obtained by the receiving end deployed on the second edge gateway and the compressed video stream container is parsed to obtain the compressed video stream, encoding parameters and frame number. Then, the compressed video stream is decoded according to the encoding parameters and frame number to obtain the frame sequence under the encoding resolution. Post-processing steps: The receiver uses a display super-resolution network to enlarge each decoded frame in the frame sequence at the encoded resolution to a preset display magnification, so as to obtain the frame sequence at the display resolution. Encapsulation and output steps: The receiving end encapsulates the frame sequence at the display resolution into a playable video file and then outputs it.

[0007] Furthermore, after obtaining the compressed video stream container in the encoding step, the method further includes: determining whether the volume of the compressed video stream container exceeds a preset target volume limit; if so, removing the corresponding video frames sequentially from the end of the compressed video stream container until the volume of the compressed video stream container is less than or equal to the preset target volume limit; wherein, while removing the corresponding video frames, the number of frames in the compressed video stream container is updated synchronously.

[0008] Furthermore, the adaptive frame extraction step, which involves performing content motion feature analysis on the video frame sequence based on a preset motion analysis model to determine the content motion feature level, specifically includes: Step 1: Calculate the motion intensity of each video frame in the video frame sequence; Step 2: Count the number of video frames whose motion intensity exceeds a preset threshold, and calculate the proportion of this number to the total number of video frames; Step 3: Determine the content motion feature level of the video frame sequence according to the ratio, and then determine the upper limit of the effective encoding frame rate, the upper limit of the number of frames to be retained by frame extraction, the target volume tolerance band, and the enhancement quota according to the content motion feature level and the system preset level table; the content motion feature level includes fully static, partially static, balanced, partially dynamic, and fully dynamic. Step 4: Extract frames from the video frame sequence based on the probe frame rate to obtain a frame-extracted sequence. Determine the number of frames to be extracted based on the upper limit of the effective coding rate and the upper limit of the number of frames to be retained. Then, perform secondary sparsification on the frame-extracted sequence based on the number of frames extracted and the preset frame-extracting rules to obtain a sparse frame sequence. Specifically, the secondary sparsification of the frame-extracted sequence based on the number of frames extracted and the preset frame-extracting rules to obtain a sparse frame sequence includes: retaining the last frame in the frame sequence, discarding video frames with motion intensity lower than a preset motion threshold, and retaining the video frame when the motion intensity of the corresponding video frame in a long static segment shows a local peak relative to the preceding and following video frames.

[0009] Furthermore, prior to the preprocessing step, the method further includes: selecting video frames to be enhanced from the sparse frame sequence according to the enhancement quota, and performing image enhancement preprocessing on the video frames to be enhanced; the image enhancement preprocessing includes performing region of interest enhancement and unsharpened mask sharpening on the corresponding video frames.

[0010] Furthermore, the encoding steps specifically include: Step 1: deriving a baseline byte rate based on the preset target volume and effective video duration, multiplying the baseline byte rate by a space complexity coefficient and a time complexity coefficient to obtain an initial encoding bitrate, then determining the encoding bitrate based on the initial encoding bitrate and a system preset minimum bitrate lower limit, and subsequently determining the variable bitrate encoding of the standard video file based on the encoding bitrate; the encoding bitrate is the maximum value between the initial encoding bitrate and the system preset minimum bitrate lower limit. Step 2: Determine the asymmetric volume tolerance band based on the content motion characteristic level, and encode each video frame in the sparse frame sequence according to the asymmetric volume tolerance band and the variable bitrate encoding of the standard video file to obtain a compressed video bitstream. At the same time, write the encoding parameters and the number of frames into the container header of the compressed video bitstream container to encapsulate it into a compressed video bitstream container; the encoding parameters include encoding width and height.

[0011] Furthermore, the post-processing steps specifically include: Step 1, deblocking each decoded frame of the frame sequence at the encoding resolution using a second preset deblocking network, and sharpening each decoded frame after deblocking using an unsharpened mask; Step 2, scaling each decoded frame in the sharpened frame sequence to a preset display magnification based on a display super-resolution network to obtain the frame sequence at the display resolution, and interpolating adjacent video frames in the frame sequence at the display resolution according to preset interpolation rules.

[0012] Furthermore, after the decoding step and the amplification step, the following steps are included: extracting a preset number of decoded frames from the current frame sequence according to preset rules for screen distortion detection; and discarding the current frame sequence when the number of distorted decoded frames exceeds the preset number of decoded frames; otherwise, proceeding to the next step; wherein, determining whether a decoded frame is distorted specifically includes: first, downsampling each decoded frame to obtain the corresponding number of grid blocks, then calculating the brightness standard deviation, high saturation pixel ratio, and difference between adjacent blocks for each grid block, and finally determining whether the corresponding sampled frame is distorted based on the brightness standard deviation, high saturation pixel ratio, and difference between adjacent grid blocks for each grid block; According to preset rules, a preset number of decoded frames are extracted from the current frame sequence. Specifically, any one of the following methods can be used: select the first N decoded frames from the current frame sequence, extract a preset number of decoded frames from the current frame sequence, or take the first and last decoded frames from the current frame sequence.

[0013] Furthermore, after the adaptive frame extraction step, the method further includes: determining whether the video frames in the sparse frame sequence are in portrait mode; if so, converting the video frames in the sparse frame sequence to landscape mode and generating a rotation flag, and then writing the rotation flag into the compressed video stream container header; wherein, when the frame height of a video frame is greater than its frame width, the corresponding video frame is in portrait mode. The decoding step further includes: when there is a rotation flag in the compressed video stream container, converting the decoded frames in the frame sequence at the encoded resolution into portrait mode.

[0014] The second objective of this invention is achieved through the following technical solution: The satellite image transmission video processing device based on an edge gateway includes a memory and a processor. The memory stores a satellite image transmission video processing program that runs on the processor. The satellite image transmission video processing program is a computer program. When the processor executes the satellite image transmission video processing program, it implements the steps of the satellite image transmission video processing method based on an edge gateway as one of the objectives of this invention.

[0015] The third objective of this invention is achieved through the following technical solution: A computer-readable storage medium storing a satellite image transmission video processing program, the satellite image transmission video processing program being a computer program, wherein when executed by a processor, the satellite image transmission video processing program implements the steps of a satellite image transmission video processing method based on an edge gateway, as one of the objectives of this invention.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention encodes video at the transmitting end using standard video encoding that meets a preset size constraint, focusing on generating a transmittable bitstream within the hard size limit. At the receiving end, the compressed video bitstream is enhanced and encapsulated into a universal video file, focusing on restoring the local or remote bitstream into a readable image. This eliminates the need for strict encoding and decoding strategies at both the transmitting and receiving ends. Furthermore, this invention dynamically adapts the compression encoding bitrate based on the detection of dynamic motion features in the video frame, thus preserving more data in dynamic scenes without exceeding the size limit and avoiding the loss of motion information. It also further saves bitrate and reduces satellite uplink bandwidth usage in static scenes. Attached Figure Description

[0017] Figure 1 This is a flowchart of the satellite image transmission video processing method based on an edge gateway provided by the present invention. Detailed Implementation

[0018] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments. It should be noted that, without conflict, the various embodiments or technical features described below can be arbitrarily combined to form new embodiments. Example 1

[0019] like Figure 1 As shown, the present invention provides a preferred embodiment of a satellite image transmission video processing method based on an edge gateway. This method is applied to an edge gateway and is deployed at both the sending and receiving ends of the edge gateway to process the video stream data of the satellite image transmission, thereby improving the data transmission quality and efficiency of the satellite image transmission.

[0020] More preferably, the transmitting and receiving ends are typically deployed on different edge gateways. The transmitting end, deployed on the local monitoring edge gateway, connects to the data acquisition device, acquires local monitoring video files, compresses them, and uploads them to the satellite via a satellite link. The receiving end, deployed on a remote or backend edge gateway, communicates with the satellite to acquire the compressed video stream from the satellite link, decodes it, and converts the video stream into a playable video file for output, thus completing satellite image transmission.

[0021] The satellite image transmission video processing method based on an edge gateway provided by this invention includes a sending end process and a receiving end process. Wherein, as... Figure 1 As shown, the sending process specifically includes: Step S11: Obtain the video frame sequence of local monitoring through the transmitter deployed on the first edge gateway. The video frame sequence includes multiple consecutive video frames.

[0022] The video frame sequence is obtained by taking screenshots of local monitoring video files or video streams. Specifically, multiple consecutive video frames are sequentially captured from the starting frame of the video file or video stream according to a preset duration; simultaneously, the original resolution of the video frames is preserved when taking screenshots. The preset duration is the upper limit of the capture window issued by the task or system configuration, used to limit the continuous video duration retained from the starting frame, and is not calculated in real time by an algorithm based on the content. More preferably, the preset duration is set to not exceed the system's preset processing time limit.

[0023] Step S12: The sending end performs content motion feature analysis on the video frame sequence based on a preset motion analysis model to determine the content motion feature level, and then extracts frames from the video frame sequence according to the content motion feature level to obtain a sparse frame sequence.

[0024] Specifically, this invention analyzes the motion characteristics of video stream content to distinguish whether the video stream's image is static or dynamic. Based on these motion characteristics, it determines the frame extraction ratio to remove redundant frames and retain only those containing important data for stable transmission. For example, when the image is static with minimal changes, fewer frames are needed for subsequent image reconstruction. Frame extraction allows for a significantly smaller output size than the target size, avoiding wasted satellite bandwidth. Conversely, when the image is dynamic with rich motion, more frames are required. In this case, the output size can exceed the target size estimate to ensure reliable decodeability.

[0025] More preferably, the content motion feature levels in this embodiment generally include fully static, partially static, balanced, partially dynamic and fully dynamic. Different content motion feature levels retain different video frames, that is, their corresponding frame extraction ratios are different, which can be configured according to actual needs.

[0026] Furthermore, step S12, which involves performing content feature analysis on the video frame sequence based on a motion analysis model to determine the content motion feature level, specifically includes: Step 1: Calculate the motion intensity of each video frame in the video frame sequence. Specifically, the motion intensity of each video frame is determined based on the grayscale difference between adjacent video frames. For example, for each video frame: At the same time, the video frame and the previous adjacent video frame are converted into grayscale images respectively. Then, the mean or mean square of the absolute difference between the grayscale values ​​of the two video frames is calculated pixel by pixel and used as the motion intensity of the video frame. Alternatively, the video frame and the preceding adjacent video frame can be divided into grid blocks of the same size. Then, the average grayscale difference or inter-block variance between corresponding grid blocks is calculated block by block. The calculation results from multiple grid blocks are aggregated (e.g., by taking the mean or maximum value) to determine the motion intensity of the video frame, which is then used as the motion intensity of the subsequent frame. Furthermore, when the video frame is the first frame, a preset initial value can be used as the motion intensity, or the inter-frame difference between the video frame and the subsequent video frame can be used as the motion intensity.

[0027] Step 2: Count the number of video frames whose motion intensity exceeds the preset threshold, and calculate the proportion of this number to the total number of video frames.

[0028] Step 3: Determine the content motion feature level of the video frame sequence based on the ratio, and then determine the upper limit of the effective encoded frame rate, the upper limit of the number of frames to be retained by frame extraction, the target volume tolerance band, and the enhancement quota based on the content motion feature level and the system preset level table. Specifically, when the ratio is not less than the first preset ratio, such as 0.80, the content motion feature level is considered to be full dynamic; when the ratio is less than the first preset ratio but not less than the second preset ratio (such as 0.55), the content motion feature level is partially dynamic; when the ratio is less than the second preset ratio but not less than the third preset ratio (such as 0.35), the content motion feature level is balanced; when the ratio is less than the third preset ratio but not less than the fourth preset ratio (minimum value), the content motion feature level is partially static; when the ratio is less than the fourth preset ratio, the content motion feature level is fully static.

[0029] Step 4: Extract frames from the video frame sequence based on the probe frame rate to obtain the extracted frame sequence. Determine the number of frames to be extracted based on the upper limit of the effective coding rate and the upper limit of the number of frames to be retained after extraction. Then, perform secondary sparsification on the extracted frame sequence based on the number of frames extracted and the preset extraction rules to obtain the sparse frame sequence.

[0030] Specifically, when extracting frames from a video frame sequence, the sequence is first extracted based on the probe frame rate, and then the extracted frames undergo secondary sparsification. The probe frame rate, used to assess motion distribution and static dominance, is typically a system-preset trial sampling frame rate higher than the final effective coding frame rate. Generally, the probe frame rate is pre-configured by the system or task, for example, by a preset multiple or fixed increment of the target coding frame rate.

[0031] During the secondary sparsification, the frame sequence is further sparsified according to the number of extracted frames and preset frame extraction rules to obtain a sparse frame sequence. Specifically, this includes: retaining the last frame in the frame sequence; discarding video frames with motion intensity below a preset motion threshold; and retaining a video frame if its motion intensity shows a local peak relative to the preceding and following video frames within a long still segment. Furthermore, retaining a video frame within a long still segment if its motion intensity shows a local peak relative to the preceding and following video frames specifically includes: when the motion intensity of several consecutive frames is below a preset stillness threshold, and the number of consecutive frames or their corresponding duration exceeds a preset still segment threshold, the continuous interval is defined as a long still segment. A local peak refers to a frame within a long still segment having a motion intensity higher than its preceding and following neighboring frames (or higher than other frames within a preset neighborhood window centered on that frame), thus forming a relative spike; that is, by forcibly retaining this frame as a sudden motion anchor point, transient motion information is avoided. Furthermore, the number of frames extracted during secondary sparsity is determined by the upper limit of the effective coding rate and the upper limit of the number of frames retained during extraction. The upper limit of the effective coding frame rate represents the upper limit of the allowed temporal sampling density (frames / second) under this motion characteristic level, used to constrain the temporal bitrate during the coding stage. The upper limit of the number of frames retained during extraction represents the hard upper limit of the maximum number of frames allowed to be retained within this segment. When performing secondary sparsity on the extracted frame sequence, the upper limits specified by the above two conditions must be satisfied simultaneously.

[0032] More preferably, to ensure the quality of video frames, the present invention also performs image enhancement preprocessing on the video frames in the sparse frame sequence. Specifically, based on the enhancement quota determined by the content motion feature level, video frames to be enhanced are selected from the sparse frame sequence, and image enhancement preprocessing is performed on these video frames. Furthermore, the image enhancement preprocessing includes performing region of interest enhancement and unsharpened mask sharpening on the corresponding video frames. The enhancement quota refers to the maximum number of frames allowed to undergo heavy enhancement processing such as region of interest enhancement and unsharpened mask sharpening at the current content motion feature level, or an upper limit calculated as a preset proportion of the total number of sparse frames, used to control enhancement overhead within the computing power and bitrate budget. That is, selecting video frames to be enhanced from the sparse frame sequence specifically includes: prioritizing frames containing moving subjects or detection targets with high motion intensity, filling the enhancement quota from high to low motion intensity; frames not included in the quota skip region of interest enhancement and unsharpened mask sharpening, only performing geometric downsampling and necessary deblocking processing.

[0033] More preferably, performing region of interest (ROI) enhancement on the corresponding video frame specifically includes: generating a ROI mask based on inter-frame motion difference, saliency detection, or target detection results; performing enhancements such as increasing local contrast or slight sharpening on the region within the mask; and weakening or skipping enhancements on the region outside the mask, thereby concentrating the encoding budget on the critical regions for interpretation. Performing mask sharpening on the corresponding video frame specifically includes: applying low-pass blur to the original frame to obtain a blurred frame; subtracting the blurred frame from the original frame to obtain a detail layer; and then superimposing the detail layer back onto the original frame according to a preset gain; both the blur radius and gain are system preset parameters.

[0034] Step S13: Perform geometric downsampling on the sparse frame sequence by the transmitting end and convert it into a frame sequence at the coding resolution, and then perform deblocking on the frame sequence at the coding resolution.

[0035] Specifically, geometric scaling, such as interpolation to reduce image size, scales the original resolution video frames to the encoding resolution specified by the encoder for subsequent encoding. Generally, interpolation reconstruction is used, calculating new pixels based on adjacent pixels to ensure geometric alignment and smooth edges in the scaled image. Deblocking is used to preprocess the original resolution video frames to eliminate spurious blocks, sensor artifacts, and ISP-induced block noise generated during data acquisition, in order to adapt to encoding, suppress high-frequency components in advance, and control the encoding bitrate. Specifically, a first preset deblocking network can be used to implement deblocking, which is a dedicated deblocking network at the transmitting end, such as an FBCNN-like network.

[0036] Step S14: The transmitting end determines the variable bitrate encoding of the standard video file according to the preset target volume and content motion feature level, and then encodes the deblocked frame sequence according to the variable bitrate encoding of the standard video file to generate a compressed video stream. The compressed video stream, encoding parameters and frame number are encapsulated into a compressed video stream container, and the compressed video stream container is uplinked through the satellite link.

[0037] The preset target volume can be obtained by configuring parameters during system startup. Furthermore, different content motion feature levels correspond to different coding rates. Therefore, step S14 also includes: First, the baseline bitrate is obtained based on the preset target volume and the effective duration of the video. Then, the baseline bitrate is multiplied by the space complexity coefficient and the time complexity coefficient to obtain the initial encoding bitrate.

[0038] Then, the encoding bitrate is determined based on the initial encoding bitrate and the system's preset minimum bitrate limit. Furthermore, the variable bitrate encoding of the standard video file is derived from this encoding bitrate. The encoding bitrate is the target bitrate control parameter for the variable bitrate encoding of the standard video file, measured in bytes per second. The encoding bitrate serves as the variable bitrate control input for the standard video encoder (such as the AV1 encoder in one embodiment), constraining the output rate of the compressed video stream. Additionally, the encoding bitrate is the maximum value between the initial encoding bitrate and the system's preset minimum bitrate limit.

[0039] Next, an asymmetric volume tolerance band is determined based on the content motion feature level. Each video frame in the sparse frame sequence is then encoded according to the asymmetric volume tolerance band and the encoding bitrate to obtain a compressed video stream. Simultaneously, the encoding parameters and the number of frames are written into the container header of the compressed video stream container to encapsulate it into a compressed video stream container. The encoding parameters include encoding width and height. Specifically, the asymmetric volume tolerance band is a pair of tolerance parameters bound according to the content motion feature level, representing the upper limit of the allowed under-bitrate and over-bitrate ratios relative to the target volume, respectively. In the static mode, a larger under-bitrate ratio is allowed, while over-bitrate ratios are strictly controlled to save satellite bandwidth; conversely, in the dynamic mode, moderate over-bitrate ratios are allowed to prioritize decodeability. Furthermore, after encoding, the target volume hard upper limit is still used as the standard; if it is exceeded, frame-by-frame pruning is performed at the end of the container.

[0040] More preferably, in step S13, after generating the compressed video stream container, it is further determined whether the volume of the compressed video stream container exceeds a preset target volume limit. If so, video frames are sequentially removed from the end of the compressed video stream container until the volume of the compressed video stream container does not exceed the preset target volume limit. When removing the corresponding video frame from the compressed video stream container, the number of frames in the compressed video stream container is also updated synchronously. Preferably, the compressed video stream container in this invention uses an AVI format IVF container.

[0041] More preferably, after step S13, the method further includes: determining whether a video frame in the sparse frame sequence is in portrait mode; if so, converting the video frame to landscape mode, for example, rotating the video frame counterclockwise by 90°. That is, rotating the video frame by a preset angle according to a preset first rotation direction. When the frame height of a video frame is greater than its frame width, the video frame is considered to be in portrait mode. By switching the display mode, the display mode of the video frame can be kept consistent with display habits. At the same time, when converting to landscape mode, a rotation flag is generated and written into the compressed video stream container header, so that the rotation flag can be used to determine whether to restore to portrait mode later.

[0042] Preferably, the receiving end process specifically includes: Step S21: Obtain the compressed video stream container through the receiver deployed on the second edge gateway and parse the compressed video stream container to obtain the compressed video stream, encoding parameters and frame count, and decode the compressed video stream according to the encoding parameters and frame count to obtain the frame sequence at the encoding resolution.

[0043] More preferably, the standard video encoding uses the AVI encoding format.

[0044] The receiving end obtains the corresponding message through the satellite link, parses the message according to the corresponding protocol format to obtain the compressed video bitstream, encoding parameters and frame count, and then decodes the compressed video bitstream according to the encoding parameters and frame count to obtain the frame sequence at the encoding resolution.

[0045] Step S22: The receiver amplifies each decoded frame in the frame sequence at the encoded resolution to a preset display magnification based on the display super-resolution network to obtain the frame sequence at the display resolution.

[0046] Specifically, the display super-resolution network can be implemented using a Real-ESRGAN-like 2X network. This is achieved by upscaling the decoded frames at the encoded resolution to the display resolution, thus ensuring accuracy for the subsequent generation of playable video files.

[0047] More preferably, to ensure the quality of the decoded frames, the present invention performs image post-processing on the decoded frames. Specifically, step S22 further includes: firstly, deblocking each decoded frame of the frame sequence at the encoded resolution using a second preset deblocking network, and then sharpening each decoded frame after deblocking using an unsharpened mask; then, based on a display super-resolution network, enlarging each decoded frame in the sharpened frame sequence to a preset display magnification, thereby obtaining the frame sequence at the display resolution. The second preset deblocking network is a dedicated deblocking network at the receiving end, which performs inference on the decoded frames after decoding to eliminate block artifacts generated during video frame compression, repair edge blurring, and optimize display quality.

[0048] In other words, this invention preprocesses video frames at the transmitting end by deblocking and sharpening them before compression. This preprocessing is applied to the uncompressed original image to adapt to encoding, suppress high-frequency components in advance, and control the encoding bitrate. At the receiving end, the decoded video frames undergo deblocking and sharpening postprocessing, which is applied to the compressed and reconstructed distorted image to eliminate coding block effects, repair edge blurring, and optimize display quality. By preventing encoding distortion at the transmitting end and repairing compression artifacts at the receiving end, the objects and roles of deblocking and sharpening differ. Therefore, the weights of the deblocking network and the parameters of the sharpening mask used in the two processes are different. Specifically, for example, the second preset deblocking network used in the receiving end's deblocking processing is implemented using a NAFNet-like network.

[0049] More preferably, when the configuration is enabled and the inter-frame motion intensity of adjacent frame pairs is within the interpolable range, an optical flow estimation network synthesizes intermediate frames to improve smoothness; interpolation is skipped for frame pairs with repetitive textures or extremely low motion to prevent snow artifacts, and low-motion frame pairs can be backed up to the nearest neighbor frame for copying. The optical flow estimation network uses a RIFE-type optical flow network. Specifically, "configuration enabled" refers to a temporal interpolation switch in the system or task configuration; interpolation is only entered when this switch is on, and skipped when it is off. The inter-frame motion intensity of adjacent frame pairs is calculated by examining the enhanced adjacent video frames at the receiver, using the same principle as the aforementioned motion intensity calculation. The interpolable range is the range enclosed by preset upper and lower thresholds for motion intensity. Below the lower threshold, interpolation yields little benefit and easily introduces artifacts; above the upper threshold, the reliability of optical flow estimation decreases. That is, interpolation is only performed when the inter-frame motion intensity of adjacent frame pairs falls within the aforementioned upper and lower threshold range.

[0050] Synthesizing intermediate frames using optical flow estimation networks specifically includes: inputting two frames, estimating bidirectional optical flow and occlusion information, and distorting and fusing the two frames according to the midpoint of time or a preset interpolation ratio to obtain intermediate frames; in one embodiment, an optical flow interpolation network can be used, but it is not limited to a specific implementation.

[0051] Repetitive textures or extremely low motion frames are defined as: frame pairs whose motion intensity is lower than a preset extremely low motion threshold; or whose block variance is lower than a preset texture threshold and whose inter-frame difference is extremely small, thus being identified as repetitive texture / static background frame pairs.

[0052] Low-motion frames can be backed up to the nearest neighbor frame for copying. This means that instead of generating an optical flow composite intermediate frame for the frame pair, the frame that is closer in time (usually the previous frame) is directly copied as an intermediate frame placeholder to avoid snow-like artifacts.

[0053] More preferably, when a landscape mode conversion occurs at the transmitting end, step S22 further includes mode conversion of the image post-processing frames, converting the landscape video frames to portrait mode to restore the original video stream display mode. Specifically, when the transmitting end detects that the frame height is greater than the frame width and rotates the portrait frame clockwise to landscape mode, it writes a rotation flag into the compressed video stream container header or accompanying parameters. Thus, when the receiving end receives the compressed video stream, it determines whether a landscape mode conversion has occurred by checking for the presence of this rotation flag. If the rotation flag is present, the landscape mode is restored to portrait mode counterclockwise; if the rotation flag is absent, no restoration is performed.

[0054] Step S23: The receiving end encapsulates the frame sequence at the display resolution into a playable video file and then outputs it.

[0055] More preferably, after decoding the compressed video stream to obtain decoded frames in step S21, it is also necessary to perform quality detection on the decoded frames. Specifically, this includes: extracting a preset number of decoded frames from the current frame sequence according to preset rules for screen distortion detection; and discarding the current frame sequence when the number of decoded frames with screen distortion exceeds the preset number of decoded frames; otherwise, proceeding to step S22.

[0056] According to preset rules, a preset number of decoded frames are extracted from the current frame sequence. Specifically, any one of the following methods can be used: select the first N decoded frames from the current frame sequence, extract a preset number of decoded frames from the current frame sequence, or take the first and last decoded frames from the current frame sequence.

[0057] Specifically, determining whether a decoded frame is distorted involves: first, downsampling each decoded frame to obtain the corresponding number of grid blocks; then, calculating the brightness standard deviation, high saturation pixel ratio, and difference between adjacent blocks for each grid block; and finally, determining whether the corresponding sampled frame is distorted based on the brightness standard deviation, high saturation pixel ratio, and difference between adjacent grid blocks for each grid block.

[0058] Similarly, after the decoded frame is amplified in step S22, a quality check is also performed on each decoded frame. In this way, the display resolution frame sequence that passes the quality check can be encoded into a browser-playable format such as H.264 for local comparison or delivery to downstream applications. The quality check is the same as above. Example 2

[0059] The satellite image transmission video processing device based on the edge gateway includes a memory and a processor. The memory stores a satellite image transmission video processing program that runs on the processor. The satellite image transmission video processing program is a computer program. When the processor executes the satellite image transmission video processing program, it performs the following steps: Video frame acquisition steps: The local monitoring video frame sequence is acquired through the transmitter deployed on the first edge gateway. The video frame sequence includes multiple consecutive video frames. Adaptive frame extraction step: The sending end performs content motion feature analysis on the video frame sequence based on a preset motion analysis model to determine the content motion feature level, and then extracts frames from the video frame sequence according to the content motion feature level to obtain a sparse frame sequence. Preprocessing steps: The sparse frame sequence is geometrically downsampled by the transmitting end and converted into a frame sequence at the coding resolution, and the frame sequence at the coding resolution is deblocked. Encoding steps: The transmitting end determines the variable bitrate encoding of the standard video file according to the preset target volume and content motion feature level. Then, the deblocked frame sequence is encoded according to the variable bitrate encoding of the standard video file to generate a compressed video stream. The compressed video stream, encoding parameters and frame number are encapsulated into a compressed video stream container, and the compressed video stream container is uplinked through the satellite link. Decoding steps: The compressed video stream container is obtained by the receiving end deployed on the second edge gateway and the compressed video stream container is parsed to obtain the compressed video stream, encoding parameters and frame number. Then, the compressed video stream is decoded according to the encoding parameters and frame number to obtain the frame sequence at the encoding resolution. Post-processing steps: The receiver uses a display super-resolution network to enlarge each decoded frame in the frame sequence at the encoded resolution to a preset display magnification, so as to obtain the frame sequence at the display resolution. Encapsulation and output steps: The receiving end encapsulates the frame sequence at the display resolution into a playable video file and then outputs it.

[0060] Furthermore, after obtaining the compressed video stream container in the encoding step, the method further includes: determining whether the volume of the compressed video stream container exceeds the preset target volume limit; if so, removing the corresponding video frames sequentially from the end of the compressed video stream container until the volume of the compressed video stream container is less than or equal to the preset target volume limit; wherein, while removing the corresponding video frames, the number of frames in the compressed video stream container is updated synchronously.

[0061] Furthermore, the adaptive frame extraction step, which involves analyzing the content motion features of the video frame sequence based on a preset motion analysis model to determine the content motion feature level, specifically includes: Step 1: Calculate the motion intensity of each video frame in the video frame sequence; Step 2: Count the number of video frames whose motion intensity exceeds a preset threshold, and calculate the proportion of this number to the total number of video frames; Step 3: Determine the content motion feature level of the video frame sequence based on the ratio, and then determine the upper limit of the effective encoding frame rate, the upper limit of the number of frames to be retained by frame extraction, the target volume tolerance band, and the enhancement quota based on the content motion feature level and the system preset level table; the content motion feature level includes fully static, partially static, balanced, partially dynamic, and fully dynamic. Step 4: Extract frames from the video frame sequence based on the probe frame rate to obtain a frame-extracted sequence. Determine the number of frames to be extracted based on the upper limit of the effective coding rate and the upper limit of the number of frames to be retained. Then, perform secondary sparsification on the frame-extracted sequence based on the number of frames extracted and the preset frame-extracting rules to obtain a sparse frame sequence. Specifically, the secondary sparsification of the frame-extracted sequence based on the number of frames extracted and the preset frame-extracting rules to obtain a sparse frame sequence includes: retaining the last frame in the frame sequence, discarding video frames with motion intensity lower than a preset motion threshold, and retaining the video frame when the motion intensity of the corresponding video frame in a long static segment shows a local peak relative to the preceding and following video frames.

[0062] Furthermore, prior to the preprocessing step, the process includes: selecting video frames to be enhanced from the sparse frame sequence according to the enhancement quota, and performing image enhancement preprocessing on the video frames to be enhanced; the image enhancement preprocessing includes performing region of interest enhancement and unsharpened mask sharpening on the corresponding video frames.

[0063] Furthermore, the encoding steps specifically include: Step 1: Calculate the baseline bitrate based on the preset target size and effective video duration, and multiply the baseline bitrate by the space complexity coefficient and time complexity coefficient to obtain the initial encoding bitrate. Then, determine the encoding bitrate based on the initial encoding bitrate and the system's preset minimum bitrate lower limit, and then determine the variable bitrate encoding of the standard video file based on the encoding bitrate; the encoding bitrate is the maximum value between the initial encoding bitrate and the system's preset minimum bitrate lower limit. Step 2: Determine the asymmetric volume tolerance band based on the content motion characteristics, and encode each video frame in the sparse frame sequence according to the asymmetric volume tolerance band and the variable bitrate encoding of the standard video file to obtain a compressed video bitstream. At the same time, write the encoding parameters and the number of frames into the container header of the compressed video bitstream container to encapsulate it into a compressed video bitstream container; the encoding parameters include encoding width and height.

[0064] Furthermore, the post-processing steps specifically include: Step 1, deblocking each decoded frame of the frame sequence at the encoding resolution using a second preset deblocking network, and sharpening each decoded frame after deblocking using an unsharpened mask; Step 2, scaling each decoded frame in the sharpened frame sequence to a preset display magnification based on a display super-resolution network to obtain the frame sequence at the display resolution, and interpolating adjacent video frames in the frame sequence at the display resolution according to preset interpolation rules.

[0065] Furthermore, after the decoding step and the amplification step, the process includes: extracting a preset number of decoded frames from the current frame sequence according to preset rules for screen distortion detection; and discarding the current frame sequence when the number of distorted decoded frames exceeds the preset number of decoded frames; otherwise, proceeding to the next step. Specifically, determining whether a decoded frame is distorted includes: first, downsampling each decoded frame to obtain the corresponding number of grid blocks; then calculating the brightness standard deviation, high saturation pixel ratio, and difference between adjacent blocks for each grid block; and finally determining whether the corresponding sampled frame is distorted based on the brightness standard deviation, high saturation pixel ratio, and difference between adjacent grid blocks for each grid block. According to preset rules, a preset number of decoded frames are extracted from the current frame sequence. Specifically, any one of the following methods can be used: select the first N decoded frames from the current frame sequence, extract a preset number of decoded frames from the current frame sequence, or take the first and last decoded frames from the current frame sequence.

[0066] Furthermore, after the adaptive frame extraction step, the process also includes: determining whether the video frames in the sparse frame sequence are in portrait mode; if so, converting the video frames in the sparse frame sequence to landscape mode and generating a rotation flag, and then writing the rotation flag into the compressed video stream container header; wherein, when the frame height of a video frame is greater than its frame width, the corresponding video frame is in portrait mode. The decoding steps also specifically include: when there is a rotation flag in the compressed video stream container, converting the decoded frames in the frame sequence at the encoded resolution to portrait mode. Example 3

[0067] A computer-readable storage medium storing a satellite image transmission video processing program thereon, the satellite image transmission video processing program being a computer program, which, when executed by a processor, performs the following steps: Video frame acquisition steps: The local monitoring video frame sequence is acquired through the transmitter deployed on the first edge gateway. The video frame sequence includes multiple consecutive video frames. Adaptive frame extraction step: The sending end performs content motion feature analysis on the video frame sequence based on a preset motion analysis model to determine the content motion feature level, and then extracts frames from the video frame sequence according to the content motion feature level to obtain a sparse frame sequence. Preprocessing steps: The sparse frame sequence is geometrically downsampled by the transmitting end and converted into a frame sequence at the coding resolution, and the frame sequence at the coding resolution is deblocked. Encoding steps: The transmitting end determines the variable bitrate encoding of the standard video file according to the preset target volume and content motion feature level. Then, the deblocked frame sequence is encoded according to the variable bitrate encoding of the standard video file to generate a compressed video stream. The compressed video stream, encoding parameters and frame number are encapsulated into a compressed video stream container, and the compressed video stream container is uplinked through the satellite link. Decoding steps: The compressed video stream container is obtained by the receiving end deployed on the second edge gateway and the compressed video stream container is parsed to obtain the compressed video stream, encoding parameters and frame number. Then, the compressed video stream is decoded according to the encoding parameters and frame number to obtain the frame sequence at the encoding resolution. Post-processing steps: The receiver uses a display super-resolution network to enlarge each decoded frame in the frame sequence at the encoded resolution to a preset display magnification, so as to obtain the frame sequence at the display resolution. Encapsulation and output steps: The receiving end encapsulates the frame sequence at the display resolution into a playable video file and then outputs it.

[0068] Furthermore, after obtaining the compressed video stream container in the encoding step, the method further includes: determining whether the volume of the compressed video stream container exceeds the preset target volume limit; if so, removing the corresponding video frames sequentially from the end of the compressed video stream container until the volume of the compressed video stream container is less than or equal to the preset target volume limit; wherein, while removing the corresponding video frames, the number of frames in the compressed video stream container is updated synchronously.

[0069] Furthermore, the adaptive frame extraction step, which involves analyzing the content motion features of the video frame sequence based on a preset motion analysis model to determine the content motion feature level, specifically includes: Step 1: Calculate the motion intensity of each video frame in the video frame sequence; Step 2: Count the number of video frames whose motion intensity exceeds a preset threshold, and calculate the proportion of this number to the total number of video frames; Step 3: Determine the content motion feature level of the video frame sequence based on the ratio, and then determine the upper limit of the effective encoding frame rate, the upper limit of the number of frames to be retained by frame extraction, the target volume tolerance band, and the enhancement quota based on the content motion feature level and the system preset level table; the content motion feature level includes fully static, partially static, balanced, partially dynamic, and fully dynamic. Step 4: Extract frames from the video frame sequence based on the probe frame rate to obtain a frame-extracted sequence. Determine the number of frames to be extracted based on the upper limit of the effective coding rate and the upper limit of the number of frames to be retained. Then, perform secondary sparsification on the frame-extracted sequence based on the number of frames extracted and the preset frame-extracting rules to obtain a sparse frame sequence. Specifically, the secondary sparsification of the frame-extracted sequence based on the number of frames extracted and the preset frame-extracting rules to obtain a sparse frame sequence includes: retaining the last frame in the frame sequence, discarding video frames with motion intensity lower than a preset motion threshold, and retaining the video frame when the motion intensity of the corresponding video frame in a long static segment shows a local peak relative to the preceding and following video frames.

[0070] Furthermore, prior to the preprocessing step, the process includes: selecting video frames to be enhanced from the sparse frame sequence according to the enhancement quota, and performing image enhancement preprocessing on the video frames to be enhanced; the image enhancement preprocessing includes performing region of interest enhancement and unsharpened mask sharpening on the corresponding video frames.

[0071] Furthermore, the encoding steps specifically include: Step 1: Calculate the baseline bitrate based on the preset target size and effective video duration, and multiply the baseline bitrate by the space complexity coefficient and time complexity coefficient to obtain the initial encoding bitrate. Then, determine the encoding bitrate based on the initial encoding bitrate and the system's preset minimum bitrate lower limit, and then determine the variable bitrate encoding of the standard video file based on the encoding bitrate; the encoding bitrate is the maximum value between the initial encoding bitrate and the system's preset minimum bitrate lower limit. Step 2: Determine the asymmetric volume tolerance band based on the content motion characteristics, and encode each video frame in the sparse frame sequence according to the asymmetric volume tolerance band and the variable bitrate encoding of the standard video file to obtain a compressed video bitstream. At the same time, write the encoding parameters and the number of frames into the container header of the compressed video bitstream container to encapsulate it into a compressed video bitstream container; the encoding parameters include encoding width and height.

[0072] Further, the post-processing steps specifically include: Step 1, deblocking each decoded frame of the frame sequence at the encoding resolution using a second preset deblocking network, and sharpening each decoded frame after deblocking using an unsharpened mask; Step 2, scaling up each decoded frame in the sharpened frame sequence to a preset display magnification based on a display super-resolution network to obtain the frame sequence at the display resolution, and interpolating adjacent video frames in the frame sequence at the display resolution according to preset interpolation rules; Furthermore, after the decoding step and the amplification step, the process includes: extracting a preset number of decoded frames from the current frame sequence according to preset rules for screen distortion detection; and discarding the current frame sequence when the number of distorted decoded frames exceeds the preset number of decoded frames; otherwise, proceeding to the next step. Specifically, determining whether a decoded frame is distorted includes: first, downsampling each decoded frame to obtain the corresponding number of grid blocks; then calculating the brightness standard deviation, high saturation pixel ratio, and difference between adjacent blocks for each grid block; and finally determining whether the corresponding sampled frame is distorted based on the brightness standard deviation, high saturation pixel ratio, and difference between adjacent grid blocks for each grid block. According to preset rules, a preset number of decoded frames are extracted from the current frame sequence. Specifically, any one of the following methods can be used: select the first N decoded frames from the current frame sequence, extract a preset number of decoded frames from the current frame sequence, or take the first and last decoded frames from the current frame sequence.

[0073] Furthermore, after the adaptive frame extraction step, the process also includes: determining whether the video frames in the sparse frame sequence are in portrait mode; if so, converting the video frames in the sparse frame sequence to landscape mode and generating a rotation flag, and then writing the rotation flag into the compressed video stream container header; wherein, when the frame height of a video frame is greater than its frame width, the corresponding video frame is in portrait mode. The decoding steps also specifically include: when there is a rotation flag in the compressed video stream container, converting the decoded frames in the frame sequence at the encoded resolution to portrait mode.

[0074] The above embodiments are merely preferred embodiments of the present invention and should not be construed as limiting the scope of protection of the present invention. Any non-substantial changes and substitutions made by those skilled in the art based on the present invention shall fall within the scope of protection claimed by the present invention.

Claims

1. A satellite image transmission video processing method based on an edge gateway, characterized in that, The satellite image transmission video processing method includes: Video frame acquisition steps: Acquire the video frame sequence of local monitoring through the transmitter deployed on the first edge gateway, the video frame sequence including multiple consecutive video frames; Adaptive frame extraction step: The transmitting end performs content motion feature analysis on the video frame sequence based on a preset motion analysis model to determine the content motion feature level, and then extracts frames from the video frame sequence according to the content motion feature level to obtain a sparse frame sequence. Preprocessing steps: The sparse frame sequence is geometrically downsampled by the transmitting end and converted into a frame sequence at the coding resolution, and the frame sequence at the coding resolution is deblocked. Encoding steps: The transmitting end determines the variable bitrate encoding of the standard video file according to the preset target volume and content motion feature level. Then, the deblocked frame sequence is encoded according to the variable bitrate encoding of the standard video file to generate a compressed video stream. The compressed video stream, encoding parameters and frame number are encapsulated into a compressed video stream container, and the compressed video stream container is uplinked through the satellite link. Decoding steps: The compressed video stream container is obtained by the receiving end deployed on the second edge gateway and the compressed video stream container is parsed to obtain the compressed video stream, encoding parameters and frame number. Then, the compressed video stream is decoded according to the encoding parameters and frame number to obtain the frame sequence under the encoding resolution. Post-processing steps: The receiver uses a display super-resolution network to enlarge each decoded frame in the frame sequence at the encoded resolution to a preset display magnification, so as to obtain the frame sequence at the display resolution. Encapsulation and output steps: The receiving end encapsulates the frame sequence at the display resolution into a playable video file and then outputs it.

2. The satellite image transmission video processing method based on an edge gateway according to claim 1, characterized in that, After obtaining the compressed video stream container in the encoding step, the method further includes: determining whether the volume of the compressed video stream container exceeds the preset target volume limit; if so, removing the corresponding video frames sequentially from the end of the compressed video stream container until the volume of the compressed video stream container is less than or equal to the preset target volume limit; wherein, while removing the corresponding video frames, the number of frames in the compressed video stream container is updated synchronously.

3. The satellite image transmission video processing method based on an edge gateway according to claim 1, characterized in that, The adaptive frame extraction step, which involves performing content motion feature analysis on the video frame sequence based on a preset motion analysis model to determine the content motion feature level, specifically includes: Step 1: Calculate the motion intensity of each video frame in the video frame sequence; Step 2: Count the number of video frames whose motion intensity exceeds a preset threshold, and calculate the proportion of this number to the total number of video frames; Step 3: Determine the content motion feature level of the video frame sequence according to the ratio, and then determine the upper limit of the effective encoding frame rate, the upper limit of the number of frames to be retained by frame extraction, the target volume tolerance band, and the enhancement quota according to the content motion feature level and the system preset level table; the content motion feature level includes fully static, partially static, balanced, partially dynamic, and fully dynamic. Step 4: Extract frames from the video frame sequence based on the probe frame rate to obtain a frame-extracted sequence. Determine the number of frames to be extracted based on the upper limit of the effective coding rate and the upper limit of the number of frames to be retained. Then, perform secondary sparsification on the frame-extracted sequence based on the number of frames extracted and the preset frame-extracting rules to obtain a sparse frame sequence. Specifically, the secondary sparsification of the frame-extracted sequence based on the number of frames extracted and the preset frame-extracting rules to obtain a sparse frame sequence includes: retaining the last frame in the frame sequence, discarding video frames with motion intensity lower than a preset motion threshold, and retaining the video frame when the motion intensity of the corresponding video frame in a long static segment shows a local peak relative to the preceding and following video frames.

4. The satellite image transmission video processing method based on an edge gateway according to claim 3, characterized in that, The preprocessing step is preceded by: selecting video frames to be enhanced from the sparse frame sequence according to the enhancement quota, and performing image enhancement preprocessing on the video frames to be enhanced; the image enhancement preprocessing includes performing region of interest enhancement and unsharpened mask sharpening on the corresponding video frames.

5. The satellite image transmission video processing method based on an edge gateway according to claim 1, characterized in that, The encoding steps specifically include: Step 1: Calculate the baseline byte rate based on the preset target volume and effective video duration, and multiply the baseline byte rate by the space complexity coefficient and time complexity coefficient to obtain the initial encoding bitrate. Then, determine the encoding bitrate based on the initial encoding bitrate and the system preset minimum bitrate lower limit, and then determine the variable bitrate encoding of the standard video file based on the encoding bitrate; the encoding bitrate is the maximum value between the initial encoding bitrate and the system preset minimum bitrate lower limit. Step 2: Determine the asymmetric volume tolerance band based on the content motion characteristic level, and encode each video frame in the sparse frame sequence according to the asymmetric volume tolerance band and the variable bitrate encoding of the standard video file to obtain a compressed video bitstream. At the same time, write the encoding parameters and the number of frames into the container header of the compressed video bitstream container to encapsulate it into a compressed video bitstream container; the encoding parameters include encoding width and height.

6. The satellite image transmission video processing method based on an edge gateway according to claim 1, characterized in that, The post-processing steps specifically include: Step 1, deblocking each decoded frame of the frame sequence at the encoding resolution using a second preset deblocking network, and sharpening each decoded frame after deblocking using an unsharpened mask; Step 2, scaling each decoded frame in the sharpened frame sequence to a preset display magnification based on a display super-resolution network to obtain the frame sequence at the display resolution, and interpolating adjacent video frames in the frame sequence at the display resolution according to preset interpolation rules.

7. The satellite image transmission video processing method based on an edge gateway according to claim 1, characterized in that, The steps following the decoding and amplification steps include: extracting a preset number of decoded frames from the current frame sequence according to preset rules for screen distortion detection; and discarding the current frame sequence when the number of distorted decoded frames exceeds the preset number of decoded frames; otherwise, proceeding to the next step. Specifically, determining whether a decoded frame is distorted includes: first, downsampling each decoded frame to obtain the corresponding number of grid blocks; then calculating the brightness standard deviation, high saturation pixel ratio, and difference between adjacent blocks for each grid block; and finally determining whether the corresponding sampled frame is distorted based on the brightness standard deviation, high saturation pixel ratio, and difference between adjacent grid blocks for each grid block. According to preset rules, a preset number of decoded frames are extracted from the current frame sequence. Specifically, any one of the following methods can be used: select the first N decoded frames from the current frame sequence, extract a preset number of decoded frames from the current frame sequence, or take the first and last decoded frames from the current frame sequence.

8. The satellite image transmission video processing method based on an edge gateway according to claim 1, characterized in that, The adaptive frame extraction step is followed by: determining whether the video frame in the sparse frame sequence is in portrait mode; if so, converting the video frame in the sparse frame sequence to landscape mode and generating a rotation flag, and then writing the rotation flag into the compressed video stream container header; wherein, when the frame height of the video frame is greater than the frame width, the corresponding video frame is in portrait mode. The decoding step further includes: when there is a rotation flag in the compressed video stream container, converting the decoded frames in the frame sequence at the encoded resolution into portrait mode.

9. A satellite image transmission video processing device based on an edge gateway, comprising a memory and a processor, wherein the memory stores a satellite image transmission video processing program that runs on the processor, characterized in that, The satellite image transmission video processing program is a computer program, and when the processor executes the satellite image transmission video processing program, it implements the steps of the satellite image transmission video processing method based on an edge gateway as described in any one of claims 1-8.

10. A computer-readable storage medium storing a satellite image transmission video processing program thereon, characterized in that, The satellite image transmission video processing program is a computer program, and when executed by a processor, it implements the steps of the satellite image transmission video processing method based on an edge gateway as described in any one of claims 1-8.

Images