A 5G-based record of interest instrument inspection storage system and method

Abstract

The application discloses a kind of based on 5G's interesting recorder inspection storage system and method, specifically including the video acquisition module of recorder, video slice encoding module, object neural network detection module, slice pre-recording buffer area module, interesting evaluation module, interesting video encoding module, video recording module, network signal estimation module, interesting video storage module, position detection module, 5G network signal strength detection module and central storage module, the application discloses a kind of based on 5G's interesting recorder inspection storage system and method can be according to video content interest degree, to automatically adjust the time size of pre-recording, guarantee key information can be entered in advance, also give consideration to video file size, improve the record integrity of pre-recording inspection, the application also according to network signal strength self-adapting priority key video is sent to cloud storage, reduce network signal problem cannot be transmitted to cloud key video recording, improve the security of recorder inspection.

Description

Technical Field

[0001] This invention relates to the field of video transmission, and more specifically to a 5G-based recorder of interest (ROI) inspection and storage system and method. Background Technology

[0002] Current dashcams and similar devices use fixed-duration buffering for pre-recording, but critical information may still be missed. Furthermore, when users travel through areas with weak network signals, crucial video recordings may fail to be transmitted to the cloud due to network issues, creating security risks. A flexible pre-recording method is needed that dynamically adjusts the pre-recording time based on the area of ​​interest and adaptively sends data to the cloud based on network signal strength, thereby improving the completeness of video recording and the security of dashcams.

[0003] In existing technology, there is a method for exchanging network signals in a vehicle dashcam (publication number CN113055855A), which includes the following steps: a signal APP collects network type and signal strength data; a first-level judgment is performed to determine whether the vehicle dashcam network signal is in one of three states: strong, medium, or weak, and the judgment data is uploaded to the operation management platform via the mobile internet; the network type and signal status of all online dashcams are centrally cached or stored and displayed on the business operation interface of the operation management platform; when the delay time for the online dashcam signal status to be transmitted back to the platform exceeds a set threshold, a second-level judgment is performed, which modifies the signal strength of the signal status to "off". This invention allows platform users to grasp the network status of the devices in a timely and intuitive manner, avoids common video monitoring failures in weak network conditions of vehicle dashcams, improves monitoring effects, reduces equipment recalls, and increases the online service life of the devices. However, it cannot automatically adjust the pre-recording time based on the interest level of the video content, ensuring that key information is not pre-recorded and that the integrity of the pre-recorded inspection records is not guaranteed. Summary of the Invention

[0004] The purpose of this invention is to provide a 5G-based system and method for inspecting and storing video recordings based on the degree of interest in the video content. This system automatically adjusts the pre-recording time based on the level of interest in the video content, thereby improving the integrity of the pre-recorded video recordings and reducing the risk of failing to transmit critical video recordings to the cloud due to network signal problems, thus enhancing the security of the recorder inspection.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A 5G-based recorder of interest (ROI) inspection and storage system includes:

[0007] Video acquisition module: used to acquire video data in real time and output YUV video frame data to the video segmentation and encoding module, the object neural network detection module and the video of interest encoding module;

[0008] Video segmentation encoding module: used to receive YUV video frames, encode and compress them to generate I-frame segments, and send them to the segmentation pre-recording buffer module;

[0009] Object neural network detection module: It receives YUV video frames from the video acquisition module, uses the YOLO deep learning object detection algorithm to identify the type, size, position and detection time of the object, and sends them to the interest evaluation module.

[0010] Segmented pre-recording buffer module: used to receive the comprehensive evaluation weights from the interest evaluation module and to receive I-frames from the video segmentation encoding module for caching;

[0011] Interest Evaluation Module: This module receives the object type, size, location, and detection time from the object neural network detection module, stores them in the object pool, calculates the comprehensive evaluation weight, sends it to the segmented pre-recording buffer module, and sends the interest evaluation correction value, size, and location of each object to the interest video encoding module.

[0012] Interest Video Encoding Module: Used to receive the interest evaluation correction value, size, position and detection time of the object from the interest evaluation module, and send the compressed video data of the object and the interest evaluation correction value to the interest video storage module;

[0013] Recording module: Based on the user's manual recording request, it obtains video cache segment data from the segmented pre-recording buffer module and receives the signal prediction value from the signal evaluation module of the network signal prediction module;

[0014] Network signal prediction module: It is used to receive the signal strength of the 5G network signal strength detection module and the position and movement direction of the position detection module to obtain the signal prediction value. The network signal prediction module sends the signal prediction value to the recording module and the evaluation of interest module.

[0015] Interest Video Storage Module: This module receives the compressed video data and interest evaluation correction values ​​of the objects from the Interest Video Encoding Module, and stores the key video segments into the central storage module according to the size of the interest evaluation correction values.

[0016] Location detection module: Determines the current location based on GPS signals and sends the result to the network signal estimation module;

[0017] 5G network signal strength detection module: used to send the signal strength to the network signal prediction module;

[0018] Central storage module: The central storage module receives and stores the video recordings from the recording module and the key video compression data from the video of interest storage module.

[0019] This invention also includes a 5G-based method for monitoring and storing data using a recorder of interest (ROI), employing the aforementioned 5G-based method.

[0020] The process of inspecting and storing a recorder of interest includes the following steps:

[0021] S1: The recorder's video acquisition module acquires video data in real time and outputs YUV video frame data to the video segmentation encoding module, the object neural network detection module, and the video of interest encoding module;

[0022] S2: The recorder's video segmentation encoding module receives YUV video frames, encodes and compresses them to generate I-frame segments, and sends them to the segmentation pre-recording buffer module.

[0023] S3: The recorder's object neural network detection module receives YUV video frames from the video acquisition module, uses the YOLO deep learning object detection algorithm to identify the processing of the YUV video frames, and obtains the type, size, and position of the identified object; the object neural network detection module identifies the object's type, size, position, and detection time and sends them to the interest evaluation module.

[0024] S4: The recorder's segmented pre-recording buffer module receives the comprehensive evaluation weights from the interest evaluation module and receives I-frames from the video segmentation encoding module for buffering, which is used for pre-recording;

[0025] S5: The recorder's interest evaluation module receives the object type, size, position, and detection time from the object neural network detection module, saves them in the object pool, calculates the comprehensive evaluation weight, sends it to the segmented pre-recording buffer module, and sends the interest evaluation correction value, size, and position of each object to the interest video encoding module.

[0026] S6: The recorder's video of interest encoding module receives the object's interest evaluation correction value, size, position, and detection time from the interest evaluation module, and sends the object's video compression data and interest evaluation correction value to the video of interest storage module.

[0027] S7: The recorder's recording module obtains video buffer segment data from the segmented pre-recording buffer module according to the user's manual recording request; and receives the signal prediction value from the signal evaluation module of the network signal estimation module;

[0028] S8: The network signal estimation module of the recorder receives the signal strength from the 5G network signal strength detection module and the position and direction of motion from the position detection module to obtain the signal estimation value. The network signal estimation module sends the signal estimation value to the recording module and the evaluation of interest module.

[0029] S9: The recorder's video of interest storage module receives the video compression data and interest evaluation correction value of the object from the video of interest encoding module, and stores the key video segments into the central storage module according to the size of the interest evaluation correction value.

[0030] S10: The recorder's position detection module locates the current position based on GPS signals, and determines the recorder's direction of motion based on historical position information for the current year. The position detection module sends the current position and direction of motion to the network signal estimation module.

[0031] S11: The recorder’s 5G network signal strength detection module scans all current 5G signal frequency bands and obtains the signal strength of each 5G channel. The 5G network signal strength detection module squares the signal strength and then sums them up to obtain the comprehensive signal strength. The 5G network signal strength detection module sends the signal strength to the network signal prediction module.

[0032] S12: The central storage module receives and stores the video recordings from the recorder's recording module and the key video compression data from the video of interest storage module.

[0033] Step S2 includes the following steps:

[0034] 2.1 The video segmentation and encoding module receives YUV video frames from the video acquisition module and performs H264 / H265 encoding and compression;

[0035] 2.2 The video segmentation encoding module receives the segmentation length from the segmentation pre-recording buffer module;

[0036] 2.3 The video segmentation encoding module encodes I-frames into segments according to their segment length;

[0037] 2.4 The video segmentation encoding module sends the I-frames to the segmentation pre-recording buffer module.

[0038] Step S4 includes the following steps:

[0039] 4.1 The segmented pre-recording buffer module receives I-frames from the video segmentation encoding module and performs segmented buffering for video pre-recording;

[0040] 4.2 The segmented pre-recording buffer module receives the comprehensive evaluation weights from the interest evaluation module;

[0041] 4.3 The fragmented pre-recording buffer module performs a comprehensive evaluation by dividing the weight by the preset quantization value and adding the minimum buffer size to obtain the buffer seconds. The square root of the buffer seconds is then used as the fragment length and the number of buffer fragments.

[0042] 4.4 The segmented pre-recording buffer module notifies the video segmented encoding module of the segment length, so as to maximize the length of the I-frame group and reduce the bitrate obtained for the same image quality;

[0043] 4.5 The fragmented pre-recording buffer module receives I-frames and caches them according to the number of fragments to be cached.

[0044] Step S5 includes the following steps:

[0045] 5.1 The object type, size, location, and detection time of the receiving object neural network detection module are stored in the object pool;

[0046] 5.2 The interest evaluation module periodically calculates the evaluation value of objects in the object pool, and calculates the evaluation weight of each object based on the object type, size and detection time;

[0047] 5.3 The interest assessment module obtains the initial weight value of the object based on the object type, divides the size by the base size to obtain the size adjustment value, divides it by the time calculation value to obtain the time adjustment value, where the time calculation value is the current time - detection time + 1. The initial weight value, the size adjustment value and the time adjustment value are multiplied together to obtain the evaluation value of the object. If the evaluation value is less than the set threshold, the object is removed from the object pool.

[0048] 5.4 The interest evaluation module accumulates the evaluation values ​​of objects in the object pool to obtain a comprehensive evaluation weight, which is then sent to the sharding pre-recording cache module;

[0049] 5.5 The signal prediction value of the network signal prediction module received by the interest assessment module is as follows: if the signal prediction value is less than the set threshold T, T is divided by the signal prediction value and rounded to obtain the correction level; if the signal prediction value is greater than the set threshold T, the correction level is one.

[0050] 5.6 The interest evaluation module multiplies the evaluation value of each object in the object pool by itself according to the correction level to obtain the evaluation correction value of each object in the object pool. When the signal is poor, the difference between the evaluation correction values ​​is amplified to allow high priority processing and obtain more bitrate.

[0051] 5.7 The Interest Assessment Module sends the Interest Assessment Correction Value, Size, and Position of each object to the Interest Video Encoding Module.

[0052] Step S6 includes the following steps:

[0053] 6.1 The video encoding module receives the interest evaluation correction value, size, position, and detection time of the object from the interest evaluation module;

[0054] 6.2 The video encoding module sorts the videos of interest according to their evaluation correction values ​​and processes them from largest to smallest to ensure that key videos are encoded first.

[0055] 6.3 The video encoding module generates the region of interest based on the object's size and position; the video encoding module uses the position to determine the center position of the object, enlarges the object size by 20%, and obtains the object's region of interest.

[0056] 6.4 The video encoding module receives YUV video frames from the video acquisition module and buffers them;

[0057] 6.5 The video of interest encoding module obtains the time interval of the segment by adding or subtracting the object segment duration and dividing by 2 based on the object detection time. The video of interest encoding module retrieves the YUV video frames from the buffer for the segment time interval.

[0058] 6.6 The video encoding module extracts a local image of interest from the video frame according to the region of interest of the object;

[0059] 6.7 The video encoding module of interest performs H264 / H265 video encoding on the captured image to obtain the compressed video data of the object;

[0060] 6.8 The video encoding module of interest sends the compressed video data of the object and the interest evaluation correction value to the video storage module of interest.

[0061] Step S7 includes the following steps:

[0062] 7.1 The recording module retrieves cached video clip data from the segmented pre-recording buffer module based on the user's manual recording request;

[0063] 7.2 The recording module stitches together the video buffer fragments into a video recording file for local recording and storage;

[0064] 7.3 The recording module receives the signal prediction value from the network signal estimation module. If the signal prediction value is greater than the set threshold T1, it is considered that the network signal is suitable for network transmission of video files in the next stage, and the recording file is transmitted and uploaded to the central storage module.

[0065] Step S8 includes the following steps:

[0066] 8.1 The network signal prediction module receives the signal strength from the 5G network signal strength detection module, performs Kalman filtering to remove noise and interference, and obtains the current strength value;

[0067] 8.2 The network signal prediction module receives the position and direction of motion of the position detection module, calculates the possible motion trajectory within 30 seconds based on the position and direction of motion, and judges the possible motion trajectory by taking 10 position points evenly.

[0068] 8.3 The network signal prediction module sequentially queries the historical values ​​of a location point based on its coordinates;

[0069] 8.4 The network signal prediction module sets the correction value for the location point to zero if the historical value is greater than or equal to the current intensity value, and sets the correction value for the location point to the current intensity value minus the historical value if the historical value is less than the current intensity value.

[0070] The 8.5 network signal prediction module takes the average of all correction values ​​as the output downlink adjustment value;

[0071] 8.6 The signal strength prediction module subtracts the downlink adjustment value from the current signal strength value to obtain the signal prediction value;

[0072] The 8.7 Network Signal Prediction Module sends the signal prediction value to the Recording Module and the Evaluation of Interest Module.

[0073] Step S8 includes the following steps:

[0074] 9.1 The video of interest storage module receives the compressed video data and interest evaluation correction value of the object from the video of interest encoding module;

[0075] 9.2 The video of interest storage module stores key video segments in sequence to the central storage module according to the size of the interest assessment correction value. In the case of poor network signal, key video information is transmitted to the central storage module first to improve the security of recorder users.

[0076] The beneficial effects of the 5G-based recorder of interest (ROI) inspection and storage system and method provided by the invention are as follows:

[0077] 1. This invention automatically adjusts the pre-recording time based on the level of interest in the video content, ensuring that key information can be recorded in advance while also taking into account the size of the video file, thereby improving the completeness of the pre-recorded inspection record.

[0078] 2. This invention adaptively prioritizes sending key videos to cloud storage based on network signal strength, reducing the risk of network signal problems preventing the transmission of key video recordings to the cloud and improving the security of recorder inspections. Attached Figure Description

[0079] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0080] Figure 1This is a flowchart of a 5G-based recorder of interest (ROI) inspection and storage system and method according to the present invention. Detailed Implementation

[0081] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0082] like Figure 1 As shown, a 5G-based recorder of interest (ROI) inspection and storage system includes: a video acquisition module 1, a video segmentation encoding module 2, an object neural network detection module 3, a segmentation pre-recording buffer module 4, an ROI evaluation module 5, an ROI video encoding module 6, a recording module 7, a network signal estimation module 8, an ROI video storage module 9, a location detection module 10, a 5G network signal strength detection module 11, and a central storage module 12.

[0083] Video Acquisition Module 1: The recorder's video acquisition module 1 acquires video data in real time and outputs YUV video frame data to the video segmentation encoding module 2, the object neural network detection module 3, and the video of interest encoding module 6.

[0084] Video Segmentation Encoding Module 2: This module receives YUV video frames and performs H264 / H265 encoding compression. It receives the segment length from the segment pre-recording buffer module 4, encodes the segments according to the segment length to generate I-frame segments, and then sends these I-frame segments to the segment pre-recording buffer module 4.

[0085] Object neural network detection module 3: The recorder's object neural network detection module 3 receives YUV video frames from video acquisition module 1, uses the YOLO deep learning object detection algorithm to process the YUV video frames, and obtains the type, size, and location of the identified objects. The object neural network detection module 3 then sends the identified object type, size, location, and detection time to the interest evaluation module 5.

[0086] Segment Pre-recording Buffer Module 4: This module receives I-frame segments from the video segment encoding module 2 and performs segment buffering for video pre-recording. The module receives the comprehensive evaluation weight from the interest evaluation module 5. It divides this weight by a preset quantization value and adds the minimum buffer size to obtain the buffer duration in seconds. The square root of this duration is used as the segment length and the number of buffered segments. The module notifies the video segment encoding module 2 of the segment length, aiming to maximize the length of the I-frame group and reduce the bitrate obtained for the same image quality. The module receives I-frame segments and buffers them according to the number of buffered segments.

[0087] Interest Evaluation Module 5: Module 5 receives the object type, size, location, and detection time from Object Detection Module 3 and stores them in the object pool. Module 5 periodically calculates the evaluation value of objects in the object pool, calculating the evaluation weight for each object based on its type, size, and detection time. Module 5 obtains the initial weight value for each object based on its type, divides its size by a baseline size to obtain a size adjustment value, divides it by (current time - detection time + 1) to obtain a time adjustment value, and multiplies the initial weight value, size adjustment value, and time adjustment value together to obtain the object's evaluation value. If the evaluation value is less than a set threshold, the object is removed from the object pool. Module 5 accumulates the evaluation values ​​of all objects in the object pool to obtain a comprehensive evaluation weight, which is then sent to the segmented pre-recording buffer module 4. Module 5 receives the signal prediction value from Network Signal Prediction Module 8. If the signal prediction value is less than a set threshold T, T is divided by the signal prediction value and rounded to obtain a correction level. If the signal prediction value is greater than the set threshold T, the correction level is one. The Interest Evaluation Module 5 multiplies the evaluation values ​​of objects in the object pool by their correction levels to obtain a corrected evaluation value for each object in the pool. When the signal is poor, the difference between the corrected evaluation values ​​is amplified, allowing higher-priority objects to be processed first and achieving a higher bitrate. The Interest Evaluation Module 5 then sends the corrected interest evaluation value, size, and position of each object to the Interest Video Encoding Module 6.

[0088] Interest-based video encoding module 6: Module 6 receives the interest assessment correction value, size, position, and detection time of the object from the interest assessment module 5. Module 6 sorts the objects according to their assessment correction values, processing them from largest to smallest to ensure that critical videos are encoded first. Module 6 generates regions of interest (ROIs) based on the object's size and position. Module 6 uses the position to determine the object's center location and enlarges the object size by 20% to obtain the ROI. Module 6 receives YUV video frames from video acquisition module 1 and buffers them. Module 6 obtains the segment time interval by adding or subtracting the object's segment duration and dividing by 2 based on the object's detection time. Module 6 retrieves YUV video frames from the buffer for each segment time interval. Module 6 extracts a local image of interest from the video frame image according to the object's ROI. Module 6 performs H.264 / H.265 video encoding on the extracted image to obtain the object's compressed video data. Module 6 sends the object's compressed video data and interest assessment correction values ​​to the interest-based video storage module 9.

[0089] Recording module 7: Based on the user's manual recording request, obtain video cache segment data from the segment pre-recording buffer module 4.

[0090] The recording module 7 splices the video buffer fragments into a video recording file for local recording and storage. The recording module 7 receives the signal prediction value from the signal evaluation module of the network signal prediction module 8. If the signal prediction value is greater than the set threshold T1, it is considered that the network signal is suitable for network transmission of video files in the next stage, and the recording file is transmitted and uploaded to the central storage module 12.

[0091] Network signal prediction module 8: Receives the signal strength from the 5G network signal strength detection module 11, performs Kalman filtering to remove noise and interference, and obtains the current strength value. Network signal prediction module 8 receives the position and direction of movement from the position detection module 10, calculates possible trajectories within 30 seconds based on the position and direction of movement, and selects 10 evenly spaced position points based on the possible trajectories. Network signal prediction module 8 sequentially queries the historical values ​​of each position point based on its coordinates. If the historical value is greater than or equal to the current strength value, the correction value for that position point is zero; if the historical value is less than the current strength value, the correction value for that position point is the current strength value minus the historical value. Network signal prediction module 8 averages all correction values ​​as the downlink adjustment value. The signal prediction value is obtained by subtracting the downlink adjustment value from the current signal strength value. Network signal prediction module 8 sends the signal prediction value to the recording module 7 and the interest assessment module 5.

[0092] Interest Video Storage Module 9: This module receives the compressed video data and interest evaluation correction values ​​from the object of the interest video encoding module 6. Based on the magnitude of the interest evaluation correction values, it stores key video segments into the central storage module 12 sequentially. In situations with poor network signal, key video information is prioritized for transmission to the central storage module 12, improving the safety of the recorder user.

[0093] Position detection module 10: The position detection module 10 determines the current position based on GPS signals and the direction of movement of the recorder based on historical position information from the current year. The position detection module 10 transmits the current position and direction of movement to the network signal estimation module 8.

[0094] 5G network signal strength detection module 11: The 5G network signal strength detection module 11 scans all current 5G signal frequency bands, obtains the signal strength of each 5G channel, squares the signal strength and sums them to obtain the comprehensive signal strength, and sends the signal strength to the network signal estimation module 8.

[0095] Central storage module 12: Central storage module 12 receives and stores the video recordings from the recording module 7 of the recorder and stores the key video compression data from the video of interest storage module 9.

[0096] This invention also includes a 5G-based method for inspecting and storing data using a recorder of interest (ROI), employing the aforementioned 5G-based ROI inspection and storage system, comprising the following steps:

[0097] 1. The video acquisition module 1 of the recorder acquires video data in real time and outputs YUV video frame data to the video segmentation encoding module 2, the object neural network detection module 3, and the video of interest encoding module 6.

[0098] 2. The video segmentation encoding module 2 of the recorder performs video encoding.

[0099] 2.1 The video segmentation and encoding module 2 receives YUV video frames from the video acquisition module 1 and performs H264 / H265 encoding and compression.

[0100] 2.2 The video segmentation encoding module 2 receives the segmentation length from the segmentation pre-recording buffer module 4.

[0101] 2.3 The video segmentation encoding module 2 encodes I-frame segments according to the segment length.

[0102] 2.4 The video segmentation encoding module 2 sends the I-frame segments to the segmentation pre-recording buffer module 4.

[0103] 3. The object neural network detection module 3 of the recorder receives YUV video frames from the video acquisition module 1, and uses the YOLO deep learning object detection algorithm to process the YUV video frames, obtaining the type, size, and location of the identified objects. The object neural network detection module 3 sends the object type, size, location, and detection time to the interest evaluation module 5.

[0104] 4. The recorder's segmented pre-recording buffer module 4 implements the video pre-recording function.

[0105] 4.1 The segmented pre-recording buffer module 4 receives I-frames from the video segmented encoding module 2 and performs segmented buffering for video pre-recording.

[0106] 4.2 The segmented pre-recording buffer module 4 receives the comprehensive evaluation weights from the interest evaluation module 5.

[0107] 4.3 The fragmented pre-recording cache module 4 performs a comprehensive evaluation weight divided by a preset quantization value and adds the minimum cache size to obtain the cache seconds. The square root of the cache seconds is then used as the fragment length and the number of cache fragments.

[0108] 4.4 The segmented pre-recording buffer module 4 notifies the video segmented encoding module 2 of the segment length, so as to maximize the length of the I-frame group and reduce the bitrate obtained with the same image quality.

[0109] 4.5 The fragmented pre-recording buffer module 4 receives I-frame fragments and caches them according to the number of fragments to be cached.

[0110] 4.6 The segmented pre-recording buffer module 4 receives the segmented video buffer segments from the recording module 7.

[0111] 5. The recorder's interest assessment module 5 assesses the level of interest in the video frame content.

[0112] 5.1 The object type, size, position, and detection time of the receiving object neural network detection module 3 are stored in the object pool.

[0113] 5.2 Interest Evaluation Module 5 periodically calculates the evaluation value of objects in the object pool, and calculates the evaluation weight of each object based on the object type, size and detection time.

[0114] 5.3 The Interest Assessment Module 5 obtains the initial weight value of the object based on its type. The initial weight value is obtained by dividing the initial weight value by the baseline weight value. The subsequent weight adjustment value is obtained by dividing the initial weight value by the baseline weight value. The time adjustment value is obtained by dividing the current time by the time adjustment value. The initial weight value, the size adjustment value, and the time adjustment value are multiplied together to obtain the object's assessment value. If the assessment value is less than a set threshold, the object is removed from the object pool.

[0115] 5.4 The interest evaluation module 5 accumulates the evaluation values ​​of the objects in the object pool to obtain a comprehensive evaluation weight, which is then sent to the sharding pre-recording cache module 4.

[0116] 5.5 The signal prediction value received by the network signal prediction module 5 from the signal prediction module 8 is calculated as follows: if the signal prediction value is less than the set threshold T, then T is divided by the signal prediction value and rounded down to the correction level; if the signal prediction value is greater than the set threshold T, the correction level is one.

[0117] 5.6 The Interest Evaluation Module 5 multiplies the evaluation values ​​of objects in the object pool according to the correction level to obtain the evaluation correction value of each object in the object pool. When the signal is poor, the difference between the evaluation correction values ​​is amplified to allow high priority processing and obtain more bitrate.

[0118] 5.7 The Interest Assessment Module 5 sends the Interest Assessment Correction Value, Size, and Position of each object to the Interest Video Encoding Module 6.

[0119] 6. The recorder's video encoding module 6 performs region of interest (ROI) video encoding.

[0120] 6.1 The video encoding module 6 receives the interest evaluation correction value, size, position and detection time of the object from the interest evaluation module 5.

[0121] 6.2 The video encoding module 6 sorts the videos according to their evaluation correction values ​​and processes them from largest to smallest to ensure that key videos are encoded first.

[0122] 6.3 The video encoding module 6 generates the region of interest (ROI) based on the object's size and position. The RIO module 6 uses the position to determine the object's center location, enlarges the object size by 20%, and obtains the RIO.

[0123] 6.4 The video encoding module 6 receives YUV video frames from the video acquisition module 1 and buffers them.

[0124] 6.5 The video of interest encoding module 6 obtains the time interval of the segment by adding or subtracting the object segment duration and dividing by 2 based on the object detection time. The video of interest encoding module 6 then retrieves the YUV video frames from the cache for the segment time interval.

[0125] 6.6 The video encoding module 6 extracts a local image of interest from the video frame according to the region of interest of the object.

[0126] 6.7 The video encoding module 6 performs H.264 / H.265 video encoding on the captured image to obtain the compressed video data of the object. 6.8 The video encoding module 6 sends the compressed video data of the object and the interest evaluation correction value to the video storage module 9.

[0127] 7. The recording module 7 of the recorder performs manual recording pre-recording processing.

[0128] 7.1 Recording module 7 obtains cached video cache fragment data from segmented pre-recording buffer module 4 according to the user's manual recording request.

[0129] 7.2 Recording Module 7 splices the video cache fragments into a video recording file for local recording and storage.

[0130] 7.3 The recording module 7 receives the signal prediction value from the signal evaluation module 8. If the signal prediction value is greater than the set threshold T1, it is considered that the network signal is suitable for network transmission of video files in the next stage, and the recording file is transmitted and uploaded to the central storage module 12.

[0131] 8. The network signal prediction module of the recorder predicts the network signal strength over a period of time.

[0132] 8.1: The network signal prediction module 8 acquires the signal strength data transmitted by the 5G network signal strength detection module 11, performs Kalman filtering to remove noise and interference, and obtains the current strength value.

[0133] 8.2 The network signal prediction module 8 receives the position and direction of movement of the position detection module 10, calculates the possible movement trajectory within 30 seconds based on the position and direction of movement, and selects 10 position points evenly to determine the possible movement trajectory.

[0134] 8.3 The network signal prediction module 8 queries the historical values ​​of the location point according to the coordinates of the location point.

[0135] 8.4 Network signal prediction module 8 If the historical value is greater than or equal to the current intensity value, the correction value of the location point is zero; if the historical value is less than the current intensity value, the correction value of the location point is the current intensity value minus the historical value.

[0136] The 8.5 Network Signal Prediction Module 8 takes the average of all correction values ​​as the output downlink adjustment value.

[0137] 8.6 Network Signal Prediction Module 8: Subtract the downlink adjustment value from the current signal strength value to obtain the signal prediction value.

[0138] 8.7 The network signal prediction module 8 sends the signal prediction value to the recording module 7 and the interest evaluation module 5.

[0139] 9. The recorder's video of interest storage module 9 stores videos of interest.

[0140] 9.1 The video storage module 9 receives the compressed video data and the interest evaluation correction value of the object from the video encoding module 6.

[0141] 9.2 The video storage module 9 stores key video segments in sequence to the central storage module 12 according to the magnitude of the interest assessment correction value. In the event of poor network signal, key video information is prioritized for transmission to the central storage module 12, thereby improving the safety of the recorder user.

[0142] 10. The recorder's position detection module 10 determines the current position based on GPS signals and the current year's position information based on historical position data to determine the recorder's direction of motion. The position detection module 10 transmits the current position and direction of motion to the network signal estimation module 8.

[0143] 11. The recorder's 5G network signal strength detection module 11 scans all current 5G signal frequency bands and obtains the signal strength of each 5G channel. The 5G network signal strength detection module 11 squares the signal strength and then sums them to obtain the comprehensive signal strength. The 5G network signal strength detection module 11 sends the signal strength to the network signal estimation module 8.

[0144] 12. The central storage module 12 receives and stores the video recordings from the recording module 7 of the recorder and the key video compression data from the video of interest storage module 9.

[0145] The specific embodiments described above further illustrate the technical problems, technical solutions, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A 5G-based recorder of interest (ROI) inspection and storage system, characterized in that, include: Video acquisition module (1): used to acquire video data in real time and output YUV video frame data to video segmentation encoding module (2), object neural network detection module (3) and video of interest encoding module (6); Video segmentation encoding module (2): used to receive YUV video frames, encode and compress them to generate I-frame segments, and send them to the segmentation pre-recording buffer module (4); Object neural network detection module (3): Used to receive YUV video frames from video acquisition module (1), and use YOLO deep learning object detection algorithm to identify the type, size, position and detection time of the object and send them to interest evaluation module (5); Segmented pre-recording buffer module (4): used to receive the comprehensive evaluation weights from the interest evaluation module (5) and to receive I-frames from the video segmentation encoding module (2) for caching; Interest Evaluation Module (5): Used to receive the object type, size, position and detection time of the object from the object neural network detection module (3) and store them in the object pool. Calculate the comprehensive evaluation weight and send it to the segmented pre-recording buffer module (4). Send the interest evaluation correction value, size and position of each object to the interest video encoding module (6). Receive the object type, size, position and detection time of the object from the object neural network detection module (3) and store them in the object pool. Interest Evaluation Module (5) periodically calculates the evaluation value of the objects in the object pool. Calculate the evaluation weight of each object based on the object type, size and detection time. Interest Evaluation Module (5) obtains the initial weight value of the object based on the object type. Divide the size by the base size to obtain the size adjustment value. Divide by (current time - detection time + 1) to obtain the time adjustment value. Multiply the initial weight value, size adjustment value and time adjustment value together to obtain the object's evaluation value. If... If the evaluation value is less than the set threshold, the object is removed from the object pool. The interest evaluation module (5) accumulates the evaluation values ​​of the objects in the object pool to obtain the comprehensive evaluation weight, and sends it to the segmented pre-recording buffer module (4). The interest evaluation module (5) receives the signal prediction value from the network signal prediction module (8). If the signal prediction value is less than the set threshold T, T is divided by the signal prediction value and rounded to obtain the correction level. If the signal prediction value is greater than the set threshold T, the correction level is one. The interest evaluation module (5) multiplies the evaluation value of the objects in the object pool according to the correction level to obtain the evaluation correction value of each object in the object pool. When the signal is poor, the difference between the evaluation correction values ​​is amplified so that high priority is processed first and more bitrate is obtained. The interest evaluation module (5) sends the interest evaluation correction value, size, and position of each object to the interest video encoding module (6). Interest Video Encoding Module (6): Used to receive the interest evaluation correction value, size, position and detection time of the object from the interest evaluation module (5), and send the video compression data of the object and the interest evaluation correction value to the interest video storage module (9); Recording module (7): Based on the user's manual recording request, it obtains video cache segment data from the segmented pre-recording buffer module (4) and receives the signal estimation value from the signal evaluation module of the network signal estimation module (8); Network signal estimation module (8): used to receive the signal strength of the 5G network signal strength detection module (11) and the position and direction of movement of the position detection module (10) to obtain the signal estimation value. The network signal estimation module (8) sends the signal estimation value to the recording module (7) and the interest evaluation module (5). Interest Video Storage Module (9): Used to receive the video compression data and interest evaluation correction value of the object of the Interest Video Encoding Module (6), and store the key video segments into the central storage module (12) according to the size of the interest evaluation correction value; Location detection module (10): Determines the current location based on GPS signals and sends it to the network signal estimation module (8); 5G network signal strength detection module (11): used to send the signal strength to the network signal estimation module (8); Central storage module (12): The central storage module (12) receives and stores the video recordings from the recording module (7) and the key video compression data from the video of interest storage module (9).

2. A 5G-based method for inspecting and storing data using a recorder of interest (ROI), employing a 5G-based ROI inspection and storage system as described in claim 1, characterized in that... The method includes the following steps: S1: The video acquisition module (1) of the recorder acquires video data in real time and outputs YUV video frame data to the video segmentation encoding module (2), the object neural network detection module (3) and the video of interest encoding module (6); S2: The video segmentation encoding module (2) of the recorder receives YUV video frames, encodes and compresses them to generate I-frame segments, and sends them to the segmentation pre-recording buffer module (4); S3: The object neural network detection module (3) of the recorder receives the YUV video frame from the video acquisition module (1), uses the YOLO deep learning object detection algorithm to identify the processing of the YUV video frame, and obtains the type, size and position of the identified object; the object neural network detection module (3) identifies the type, size, position and detection time of the object and sends them to the interest evaluation module (5); S4: The pre-recording buffer module (4) of the recorder receives the comprehensive evaluation weight from the interest evaluation module (5) and receives the I-frame group of video segmentation encoding module (2) for buffering and pre-recording; S5: The recorder's interest evaluation module (5) receives the object type, size, position and detection time of the object from the object neural network detection module (3) and saves them in the object pool. It calculates the comprehensive evaluation weight and sends it to the segmented pre-recording buffer module (4). It sends the interest evaluation correction value, size and position of each object to the interest video encoding module (6). S6: The video of interest encoding module (6) of the recorder receives the object's interest evaluation correction value, size, position and detection time from the object's interest evaluation module (5), and sends the object's video compression data and interest evaluation correction value to the video of interest storage module (9); S7: The recording module (7) of the recorder obtains video buffer segment data from the segmented pre-recording buffer module (4) according to the user's manual recording request; and receives the signal prediction value from the signal evaluation module of the network signal estimation module (8); S8: The network signal estimation module (8) of the recorder receives the signal strength of the 5G network signal strength detection module (11) and the position and direction of motion of the position detection module (10) to obtain the signal estimation value. The network signal estimation module (8) sends the signal estimation value to the recording module (7) and the interest evaluation module (5). S9: The video of interest storage module (9) of the recorder receives the video compression data and interest evaluation correction value of the object from the video of interest encoding module (6), and stores the key video segments into the central storage module (12) according to the size of the interest evaluation correction value. S10: The recorder's position detection module (10) locates the current position based on the GPS signal and determines the recorder's movement direction based on the historical position location information of the current year. The position detection module (10) sends the current position and movement direction to the network signal estimation module (8). S11: The 5G network signal strength detection module (11) of the recorder scans all current 5G signal frequency bands and obtains the signal strength of each 5G channel. The 5G network signal strength detection module (11) squares the signal strength and then accumulates it to obtain the comprehensive signal strength. The 5G network signal strength detection module (11) sends the signal strength to the network signal estimation module (8). S12: The central storage module (12) receives the video recording and storage of the recording module (7) of the recorder and the video of interest storage module. (9) Storage of key video compression data.

3. The method for monitoring and storing data using a 5G-based recorder of interest (ROI) according to claim 2, characterized in that, The Step S2 includes the following steps: 2.1 Video Segmentation Encoding Module (2) Receives video acquisition module (1) Performs H264 / H265 encoding compression on YUV video frames; 2.2 The video segmentation encoding module (2) receives the segmentation length from the segmentation pre-recording buffer module (4); 2.3 Video Segmentation Encoding Module (2) Encodes I-frame segments according to segment length; 2.4 The video segmentation encoding module (2) sends the I-frames to the segmentation pre-recording buffer module (4).

4. The method for monitoring and storing data using a 5G-based recorder of interest (ROI) according to claim 2, characterized in that, The Step S4 includes the following steps: 4.1 The segmented pre-recording buffer module (4) receives I-frames from the video segmentation encoding module (2) and performs segmented buffering for video pre-recording; 4.2 The segmented pre-recording buffer module (4) receives the comprehensive evaluation weight from the interest evaluation module (5); 4.3 The segmented pre-recording buffer module (4) performs a comprehensive evaluation weight divided by the preset quantization value and adds the minimum buffer size to obtain the buffer seconds. The square root of the buffer seconds is used as the segment length and the number of buffer segments. 4.4 The segmented pre-recording buffer module (4) notifies the video segmented encoding module (2) of the segment length, and tries to increase the length of the I-frame group to reduce the bit rate obtained with the same picture quality; 4.5 Segmented Pre-recording Buffer Module (4) Receives I-frame segments and caches them according to the number of cached segments.

5. The method for monitoring and storing data using a 5G-based recorder of interest (ROI) according to claim 2, characterized in that, The Step S6 includes the following steps: 6.1 The video encoding module (6) receives the interest evaluation correction value, size, position and detection time of the object from the interest evaluation module (5); 6.2 The video encoding module (6) sorts the videos according to the evaluation correction values ​​and processes them from largest to smallest to ensure that key videos are encoded first; 6.3 The video encoding module (6) generates the region of interest based on the object size and position; the video encoding module (6) uses the position to determine the center position of the object, enlarges the object size by 20%, and obtains the region of interest of the object; 6.4 The video encoding module (6) receives YUV video frames from the video acquisition module (1) and buffers them; 6.5 The video encoding module (6) obtains the time interval of the segment by adding or subtracting the object segment duration and dividing by 2 based on the object detection interest time. The video encoding module (6) retrieves the YUV video frames from the cache for the segment time interval. 6.6 The video encoding module (6) extracts local images of interest from the video frame according to the region of the object; 6.7 The video encoding module (6) performs H264 / H265 video encoding on the captured image to obtain the compressed video data of the object; 6.8 The video encoding module (6) sends the video compression data of the object and the interest evaluation correction value to the video storage module (9).

6. The method for 5G-based recorder of interest (ROI) inspection and storage according to claim 2, characterized in that, The 7.1 The recording module (7) obtains the cached video cache fragment data from the segmented pre-recording buffer module (4) according to the user's manual recording request; 7.2 The recording module (7) splices the video cache fragment data into a video recording file for local recording and storage; 7.3 The recording module (7) receives the signal prediction value from the signal evaluation module of the network signal prediction module (8). If the signal prediction value is greater than the set threshold T1, it is considered that the network signal is suitable for network transmission of video files in the next stage, and the recording file is transmitted and uploaded to the central storage module (12).

7. A 5G-based method for inspecting and storing data using a recorder of interest (ROI), as described in claim 2, is characterized in that... The Step S8 includes the following steps: 8.1 The network signal prediction module (8) receives the signal strength from the 5G network signal strength detection module (11), performs Kalman filtering to remove noise and interference, and obtains the current strength value; 8.2 The network signal prediction module (8) receives the position and direction of movement of the position detection module (10), calculates the possible movement trajectory within 30 seconds based on the position and direction of movement, and judges the possible movement trajectory by taking 10 position points evenly. 8.3 The network signal prediction module (8) queries the historical values ​​of the location point according to the coordinates of the location point; 8.4 Network signal prediction module (8) If the historical value is greater than or equal to the current intensity value, the correction value of the location point is zero; if the historical value is less than the current intensity value, the correction value of the location point is the current intensity value minus the historical value. 8.5 The network signal prediction module (8) takes the average of all correction values ​​as the output downlink adjustment value; 8.6 Network signal prediction module (8) Subtract the downlink adjustment value from the current signal strength value to obtain the signal prediction value; 8.7 The network signal prediction module (8) sends the signal prediction value to the recording module (7) and the interest evaluation module (5).

8. A 5G-based method for inspecting and storing data using a recorder of interest (ROI), as described in claim 2, is characterized in that... The Step S8 includes the following steps: 9.1 The video of interest storage module (9) receives the video compression data and interest evaluation correction value of the object from the video of interest encoding module (6); 9.2 The video storage module (9) stores key video segments in the central storage module (12) according to the size of the interest assessment correction value. In the case of poor network signal, the key video information is transmitted to the central storage module (12) first to improve the security of the recorder user.

Images