A video transmission method and system for a law enforcement recorder based on an intelligent police cloud platform

Abstract

The present application relates to law enforcement record instrument video transmission technical field, specifically, it relates to a kind of video transmission method and system for law enforcement record instrument based on intelligent police cloud platform.The method comprises the following steps: S1, the video data photographed by law enforcement record instrument is collected, and the network signal strength connected with intelligent police cloud platform of law enforcement record instrument is detected simultaneously;S2, brake light identification model is established, then according to brake light identification model in video data vehicle is divided into to-be-detected vehicle and running vehicle, and the image extraction of license plate and portrait of the nearest to-be-detected vehicle is carried out;According to network signal strength, transmission strategy is intelligently adjusted, whether in the high signal value area of network signal good, or in the medium, low signal value area of weak signal, the effective transmission of video data can be ensured, high signal real-time transmission high-definition video, medium signal transmission compressed video clip, low signal adopts the mode of timing image extraction combined with compressed video clip.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of video transmission of law enforcement recorders, in particular to a video transmission method and system for law enforcement recorders based on an intelligent police cloud platform. BACKGROUND

[0002] As an important device in the law enforcement process, the law enforcement recorder plays a key role in recording the scene, ensuring fair law enforcement, and collecting evidence. Its video transmission function aims to transmit the images and video data of the law enforcement scene to the relevant platform in real time or subsequently, so as to store, analyze and process them later.

[0003] Currently, when the law enforcement recorder transmits video to the intelligent police cloud platform, it often lacks flexible coping strategies when facing different network signal strengths. In areas with good network signals, it may be able to transmit video normally, but in areas with unstable or weak signals, such as remote areas, underground sites or network congestion periods, it is prone to video transmission interruption, lag or reduced image quality, which seriously affects the timely transmission and integrity of law enforcement information. In addition, law enforcement personnel need to input the information of the vehicle to be detected in the intelligent police cloud platform to obtain the complete information of the vehicle to be detected, which prolongs the detection efficiency of the vehicle and easily causes road congestion. Therefore, a video transmission method and system for law enforcement recorders based on an intelligent police cloud platform are proposed. SUMMARY

[0004] The present application aims to provide a video transmission method and system for law enforcement recorders based on an intelligent police cloud platform to solve the problems raised in the background.

[0005] To achieve the above technical problems, one of the purposes of the present application is to provide a video transmission method for law enforcement recorders based on an intelligent police cloud platform, comprising the following steps:

[0006] S1, collecting video data captured by the law enforcement recorder, and detecting the network signal strength connected between the law enforcement recorder and the intelligent police cloud platform;

[0007] S2, establishing a brake light recognition model, then dividing the vehicles in the video data into vehicles to be detected and vehicles in motion according to the brake light recognition model, and extracting the license plate and portrait image of the nearest vehicle to be detected;

[0008] S3, when the portrait image is extracted, detecting the start mark in the video data according to the time node of extraction, setting the end time threshold, detecting the end mark according to the extracted new portrait image and the time threshold, and then segmenting the video clip in the video data according to the time node of the detection start mark and the detection end mark;

[0009] S4, according to the network signal strength, the video segment is cached, and the video segment is compressed and packaged, and then the video segment and the image of the license plate image and the portrait image are transmitted and distributed according to priority;

[0010] S5, the intelligent police cloud platform receives the license plate image and the portrait image and the video segment, and then identifies the identity information, and feeds back the identified identity information to the corresponding user of the law enforcement recorder according to the sending source.

[0011] As a further improvement of the technical solution, the S1 establishes a network data transmission connection between the law enforcement recorder and the intelligent police cloud platform, and then the law enforcement recorder continuously supplies the video data captured by the camera to the intelligent police cloud platform for collection.

[0012] As a further improvement of the technical solution, the S1 divides the network signal strength into high, medium and low;

[0013] The network signal strength high represents that the network speed can meet the real-time transmission of high-quality video data;

[0014] The network signal strength medium represents that the network speed can meet the real-time transmission of low-quality video data, but not high-quality video data;

[0015] The network signal strength low represents that the network speed does not meet the real-time transmission of video data.

[0016] As a further improvement of the technical solution, the steps of the S2 are as follows:

[0017] S2.1, collect vehicle data in the network, and extract image data of the brake light according to the vehicle data, and then summarize and establish a brake light recognition model according to the image data;

[0018] S2.2, the video data captured by the law enforcement recorder is combined with the brake light recognition model to detect and analyze the vehicle, and the vehicle is extracted in the video data, and then the extracted vehicle is identified by the brake light recognition model, when the brake light recognition model identifies that the brake light of the vehicle is on, the vehicle is defined as a to-be-detected vehicle, otherwise, when the brake light recognition model identifies that the brake light of the vehicle is not on, the vehicle is defined as a to-be-detected vehicle;

[0019] S2.3, according to the real-time video data, the distance of the to-be-detected vehicle is analyzed, so as to determine the to-be-detected vehicle closest to the law enforcement recorder, and then the license plate and the driver of the to-be-detected vehicle are identified in the video data, when the license plate and the driver of the to-be-detected vehicle are identified in the video data, the license plate image and the portrait image of the driver are extracted.

[0020] As a further improvement of the technical solution, the S2.3 is used to extract the license plate image and the portrait image of the driver when no same segment can simultaneously display the license plate and the driver, and the segment is extracted as the image of the license plate and the driver when the license plate and the driver appear in the same segment.

[0021] Meanwhile, when the image of the license plate and the driver appearing in the same segment has been extracted, the continuous extraction is stopped, and the extracted license plate image and the portrait image of the driver are overlaid.

[0022] As a further improvement of the technical solution, the steps of the S3 are as follows:

[0023] S3.1, when the portrait image is extracted, a detection start mark is marked in the video data according to the extracted time node;

[0024] S3.2, a time threshold is set, when the detection start mark is marked in the video data, the time threshold is combined to calculate when the portrait image corresponding to the detection start mark does not appear in the video data, and when the time threshold is exceeded, the portrait image corresponding to the detection start mark does not appear in the video data, that is, a detection end mark is marked in the video data;

[0025] S3.3, when a new portrait image is extracted, a detection end mark is marked according to the extracted time node, and a new detection start is marked at the same time;

[0026] S3.4, when the detection end mark is marked by S3.2 and S3.3, the video segment is segmented in the video data according to the time nodes of the detection start mark and the detection end mark, and the video segment dedicated to the portrait image is obtained.

[0027] As a further improvement of the technical solution, the steps of the S4 are as follows:

[0028] S4.1, the video segment is cached according to the network signal strength, when the network signal strength is high, the video segment is not saved, when the network signal strength is medium or the network signal strength is low, the video segment is cached and compressed and packaged;

[0029] S4.2, the cached video segment is supplemented and sent to the intelligent police service cloud platform by using the redundant network space, and is deleted after being sent;

[0030] S4.3, when the network signal strength is in a high signal value, the law enforcement recorder transmits the video segment to the intelligent police service cloud platform in real time with high quality;

[0031] When the network signal strength is in a medium signal value, the law enforcement recorder transmits the compressed and packaged video segment to the intelligent police service cloud platform;

[0032] When the network signal strength is in a low signal value, the law enforcement recorder extracts a high-quality image from the video clip in a timing manner, and then transmits the extracted high-quality image and the compressed and packaged video clip to the intelligent police cloud platform, and the transmission priority of the high-quality image is higher than that of the compressed and packaged video clip.

[0033] As a further improvement of the technical solution, the S4 preferentially transmits the images of the license plate and the portrait extracted by the S2, and then transmits the video clip and the high-quality image according to the signal strength.

[0034] The second purpose of the present application is to provide a video transmission system for a law enforcement recorder based on an intelligent police cloud platform, which comprises a video transmission method for a law enforcement recorder based on an intelligent police cloud platform according to any one of the above, and comprises a video acquisition unit, a video clip segmentation unit and a feedback sending unit.

[0035] The video acquisition unit is used for acquiring video data shot by the law enforcement recorder, and detecting the network signal strength of the connection between the law enforcement recorder and the intelligent police cloud platform.

[0036] The video clip segmentation unit is used for establishing a brake light recognition model, and then dividing vehicles into to-be-detected vehicles and running vehicles in the video data according to the brake light recognition model, and simultaneously extracting images of license plates and portraits, and then marking detection start and detection end, so as to complete video clip segmentation.

[0037] The feedback sending unit is used for priority transmission allocation according to the network signal strength, and then the intelligent police cloud platform feeds back the identity information according to the received data.

[0038] Compared with the prior art, the present application has the following advantages:

[0039] 1. A video transmission method and system for a law enforcement recorder based on an intelligent police cloud platform, which intelligently adjusts the transmission strategy according to the network signal strength, and can ensure the effective transmission of video data in both high signal value areas with good network signals and medium and low signal value areas with weak signals. When the signal is high, high-quality video is transmitted in real time, when the signal is medium, compressed video clips are transmitted, and when the signal is low, timing image extraction combined with compressed video clips are used, and key images are preferentially transmitted. Even if the network condition is unstable, the key information of license plates and portrait images can be transmitted as quickly as possible, which provides support for the rapid decision of the command center, avoids information delay caused by network problems, and improves the overall law enforcement response speed.

[0040] 2. A video transmission method and system for a law enforcement recording instrument based on an intelligent police cloud platform, which utilizes the excess network space to supplement the transmission of cached video clips, avoids the waste of network resources, dynamically adjusts the transmission content and priority according to the network state, ensures the transmission of important information such as high-quality license plate and portrait images under the condition of limited network bandwidth, makes the network resources be reasonably allocated and maximally utilized, effectively avoids network congestion, ensures that the key data of each law enforcement point can be transmitted in time, and improves the operation efficiency of the entire law enforcement system. BRIEF DESCRIPTION OF DRAWINGS

[0041] Figure 1 is a flowchart of the whole process of the present application;

[0042] Figure 2 is a flowchart of the process of establishing a brake light recognition model according to image data in the present application;

[0043] Figure 3 is a flowchart of the process of setting an end time threshold in the present application;

[0044] Figure 4 is a flowchart of the process of caching video clips according to network signal strength in the present application;

[0045] Figure 5 is a structural schematic diagram of the video acquisition unit in the present application.

[0046] The meanings of the various reference numerals in the drawings are as follows:

[0047] 10, video acquisition unit; 20, video clip segmentation unit; 30, feedback transmission unit. DETAILED DESCRIPTION

[0048] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0049] As shown in Figures 1-5 One of the objects of the present application is to provide a video transmission method for a law enforcement recording instrument based on an intelligent police cloud platform, which comprises the following steps:

[0050] S1, acquiring video data shot by a law enforcement recording instrument, and detecting the network signal strength of the connection between the law enforcement recording instrument and an intelligent police cloud platform;

[0051] S1 establishes a network data transmission connection between the law enforcement recorder and the intelligent police cloud platform, and then the law enforcement recorder continuously supplies video data captured by the camera to the intelligent police cloud platform for collection. The specific steps are as follows:

[0052] Video data collection: After the law enforcement recorder starts the camera, the camera starts continuous video shooting according to the preset parameters (such as resolution, frame rate, etc.), converts the optical signal captured by the camera into digital video signal, and performs preliminary processing such as video encoding preprocessing (including color space conversion, denoising, etc.).

[0053] Video transmission: The law enforcement recorder encapsulates the encoded and compressed video data into network data packets, which contain video data, related metadata (such as timestamp, location information, device number, etc.), and protocol header information required for network transmission. Through the established network connection (Wi-Fi or mobile network), the data packets are sent to the intelligent police cloud platform. The network transmission protocol used is usually related protocols in the TCP / IP protocol stack, such as HTTP (Hypertext Transfer Protocol) or RTSP (Real-Time Streaming Protocol) etc. For real-time video transmission, RTSP protocol is used.

[0054] S1 divides the network signal strength into high, medium and low;

[0055] High network signal strength means that the network speed can meet the real-time transmission of high-quality video data;

[0056] The camera performs video collection at the highest resolution (e.g. 1920x1080 or higher) and the highest frame rate (e.g. 30fps or higher) to obtain the clearest and smoothest video picture.

[0057] The law enforcement recorder transmits the encoded high-quality video data to the intelligent police cloud platform at a stable high speed through the network, using real-time streaming protocol (such as RTSP) or dynamic adaptive streaming over HTTP technology (such as HLS or DASH), to ensure the real-time and continuity of video data.

[0058] Medium network signal strength means that the network speed can meet the real-time transmission of low-quality video data, but not high-quality video data;

[0059] The law enforcement recorder automatically adjusts the camera's collection parameters, reduces the resolution (e.g. to 1280x720) and frame rate (e.g. to 15fps), to reduce the amount of video data.

[0060] The LPR transmits compressed and degraded video data at a lower rate. It also uses protocols suitable for real-time transmission, such as RTSP or HTTP streaming protocols, but allocates bandwidth reasonably according to current network conditions and video data volume.

[0061] Low network signal strength represents that network speed does not meet real-time transmission of video data;

[0062] Turn on the video data buffering mechanism to temporarily store the video data that has been collected but not yet transmitted in the local cache area (e.g. using high-speed flash memory).

[0063] The LPR has a built-in network signal monitoring module that scans and obtains the signal strength indicators of the current connected network in real time, achieved through communication interaction with network base stations or Wi-Fi access points, such as receiving power levels, signal-to-noise ratios, and other parameters of wireless signals.

[0064] The obtained signal strength values are classified and judged according to the preset threshold range. Assuming that the signal strength RSSI (Received Signal Strength Indication) is used as an example, RSSI ≥ -60 dBm can be set as high signal strength, -60 dBm > RSSI ≥ -80 dBm as medium signal strength, and RSSI < -80 dBm as low signal strength. The actual threshold can be adjusted and optimized according to the specific network environment and device performance.

[0065] Once the network signal strength returns to the medium or above level, the LPR uploads the buffered video data to the cloud platform in a certain priority order.

[0066] S2, establish a brake light recognition model, then divide the vehicles in the video data into to-be-detected vehicles and running vehicles according to the brake light recognition model, and extract the license plate and image of the nearest to-be-detected vehicle;

[0067] The steps of S2 are as follows:

[0068] S2.1, collect vehicle data in the network, and extract image data of the brake light according to the vehicle data, then establish a brake light recognition model according to the image data, the specific steps are as follows:

[0069] Vehicle data collection: use traffic monitoring video data, vehicle driving record video and other sources on the network to collect image and video data containing vehicles, which cover different scenes (such as day, night, sunny, rainy, etc.), different vehicle types and driving states;

[0070] Brake light image data extraction: For the collected vehicle video data, use video frame extraction technology to extract video frame images every certain time (e.g. every second), analyze each frame image using image processing algorithms, detect the position of the vehicle and the area of the brake light, and use color feature-based methods, as brake lights usually have unique red color features, for example, set the red color range in the RGB space to (150, 0, 0) to (255, 50, 50), by traversing the image pixels, filter out the area that meets the color range as the possible brake light area, further verify and confirm the detected brake light area, such as through shape features (brake lights are generally rectangular or approximately rectangular), brightness features (brake lights are brighter when lit), etc. If these features are met, mark the frame image as a brake light lit image and store it in a dedicated brake light image data set;

[0071] Brake light recognition model establishment: Label the extracted brake light image data set, mark the brake light lit image as "positive sample", extract effective features for model training based on the labeled image data, then select the most contributing features to brake light recognition through feature selection algorithms such as principal component analysis (PCA), mutual information, etc. to reduce feature dimension, improve model training efficiency and accuracy;

[0072] According to the task characteristics and data size, select appropriate machine learning or deep learning models, for example, input the labeled image data into the network for training, adjust the network parameters through backpropagation algorithm during training, so that the model can learn the feature patterns of brake lights, thereby accurately classifying the brake light lit state. Finally, use a part of the image data that did not participate in training as a test set to evaluate the trained brake light recognition model. According to the evaluation results, analyze the shortcomings of the model, such as overfitting or underfitting, etc. If there is overfitting, methods such as increasing data augmentation (such as random rotation, scaling, cropping, etc.), adding regularization terms (such as L1, L2 regularization) can be used for optimization; If there is underfitting, try to increase the complexity of the model (such as increasing the number of network layers, increasing the number of neurons, etc.), adjust the learning rate, etc. Continuously optimize the model until the performance indicators are satisfactory.

[0073] S2.2, combine the video data captured by the law enforcement recorder with the brake light recognition model to analyze the vehicle, extract the vehicle in the video data, then identify the extracted vehicle through the brake light recognition model, when the brake light recognition model identifies that the vehicle's brake light is lit, define the vehicle as a vehicle to be detected, otherwise, when the brake light recognition model identifies that the vehicle's brake light is not lit, define the vehicle as a vehicle to be detected, the specific steps are as follows:

[0074] Video frame decomposition: The video taken by the law enforcement recorder is decomposed into continuous image frames at a fixed frame rate (e.g. 30 frames per second), which can be achieved by functions in a video processing library (such as OpenCV);

[0075] Vehicle detection: A target detection algorithm is used to detect vehicles in each image frame, and the formula is as follows:

[0076] B ij = ( x ij1 ,y ij1 ,x ij2 ,y ij2 )

[0077] where j is the serial number of the detected vehicle in the current frame, i is the frame, B ij is the left upper corner coordinate X ij1 , y ij1 and the right lower corner coordinate X ij2 , y ij2 of the rectangular bounding box of vehicle j in frame i;

[0078] A confidence threshold T is set (e.g. 0.5 according to experience) to filter out reliable vehicle detection results, and when C ij ≥ T, C ij is the confidence score, and the detected vehicle is considered valid and is extracted from the image;

[0079] The extracted vehicle image is then preprocessed, and the same feature extraction method as used in the brake light recognition model training (such as the color feature and texture feature mentioned above) is used to extract features from the vehicle image to obtain a feature vector, and the feature vector is input into the trained brake light recognition model;

[0080] Let the brake light recognition model output the probability of the brake light of the vehicle being on as P ij ;

[0081] P ij > 0.5, vehicle j in frame i is defined as "to-be-detected vehicle", indicating that the brake light is on;

[0082] P ij ≤ 0.5, vehicle j in frame i is defined as "non-to-be-detected vehicle", indicating that the brake light is not on.

[0083] S2.3, according to the real-time video data, the distance of the vehicle to be detected is analyzed, so as to determine the distance of the vehicle to be detected closest to the law enforcement recorder, and then the license plate and the driver of the vehicle to be detected are identified in the video data, when the license plate and the driver of the vehicle to be detected are identified in the video data, the license plate image and the portrait image of the driver are extracted, and the specific steps are as follows:

[0084] Target detection and tracking: the target detection algorithm is used for continuous detection of vehicles in the video, for each frame of video image, the detection box information of the vehicle is obtained, and the target tracking algorithm is used for tracking the detected vehicle, so as to assign a unique identifier to each vehicle, so as to associate the same vehicle in continuous frames;

[0085] Distance estimation: assuming that the camera parameters of the law enforcement recorder are known, such as focal length f, pixel size p, etc., for the detected vehicle, the pixel height h of the vehicle in the image is measured ij According to the principle of similar triangles, the distance of the vehicle to the law enforcement recorder can be estimated, assuming that the actual height of the vehicle is H, for common vehicle types, the average height can be used as prior knowledge, for example, the average height of a small car is about 1.5 meters, and the formula is as follows:

[0086] Solving

[0087] In each frame, all the distance values d of the vehicles marked as vehicles to be detected (i.e. vehicles with brake lights on) are traversed ij The vehicle with the minimum distance value is found, which is recorded as the nearest vehicle to be detected;

[0088] License plate recognition: for the determined nearest vehicle to be detected, the license plate region in the video frame is positioned, and once the license plate region is detected, the license plate image is extracted.

[0089] Driver identification: the face recognition technology is used to identify the driver of the vehicle, and when the driver's face is detected, the license plate image is extracted.

[0090] S2.3 When extracting the image, when no same segment can display the license plate and the driver at the same time, the license plate image and the portrait image of the driver are extracted respectively, when the license plate and the driver appear in the same segment, the segment is extracted as the image of the license plate and the driver;

[0091] At the same time, when the image of the license plate and the driver appearing in the same segment has been extracted, the extraction is stopped, and the extracted license plate image and the portrait image of the driver are covered.

[0092] S3, when the portrait image is extracted, the detection start mark is marked in the video data according to the time node of the extracted time node, and the end time threshold is set, the detection end mark is marked according to the new portrait image and the time threshold, and then the video segment is segmented in the video data according to the time node of the detection start mark and the detection end mark;

[0093] The steps of S3 are as follows:

[0094] S3.1, when the portrait image is extracted, the detection start mark is marked in the video data according to the time node of the extracted time node;

[0095] S3.2, set the end time threshold, which can be determined according to actual demand and scene, for example, it can be set to 60 seconds (indicating that if the corresponding portrait image does not appear again within 60 seconds after the detection starts, the detection is considered to end), when the video data is marked with the detection start mark, the video data does not appear the portrait image corresponding to the detection start mark, and the end time threshold is combined to start calculation, when the end time threshold is exceeded, the portrait image corresponding to the detection start mark still does not appear, that is, the detection end mark is marked in the video data;

[0096] S3.3, when the new portrait image is extracted, the detection end mark is marked according to the time node of the extracted time node, and the new detection start is marked at the same time;

[0097] S3.4, when S3.2 and S3.3 mark the detection end mark, the video segment is segmented in the video data according to the time node of the detection start mark and the detection end mark, and the video segment dedicated to the portrait image is obtained.

[0098] S4, according to the network signal strength, the video segment is cached, and the video segment is compressed and packaged, and then the video segment and the image of the license plate image and the portrait image are transmitted and allocated according to the priority;

[0099] The steps of S4 are as follows:

[0100] S4.1, according to the network signal strength, the video segment is cached, when the network signal strength is high, that is, the video segment is not saved, when the network signal strength is medium and the network signal strength is low, the video segment is cached and compressed and packaged;

[0101] S4.2, use the redundant network space to supplement the cached video segment to the intelligent police cloud platform, and delete after sending;

[0102] Evaluate the current network usage, and determine whether there is redundant network space for video segment transmission. For example, if the current network bandwidth utilization is less than 70%, it is considered that there is redundant network space;

[0103] deleting the buffered video segments that have been completely supplemented and transmitted;

[0104] S4.3, when the network signal strength is at a high signal value, the law enforcement recorder transmits the video segments to the intelligent police service cloud platform in high quality in real time;

[0105] When the network signal strength is at a medium signal value, the law enforcement recorder transmits the compressed and packaged video segments to the intelligent police service cloud platform;

[0106] When the network signal strength is at a low signal value, the law enforcement recorder extracts timed images in high quality from the video segments, and then transmits the extracted high-quality images combined with the compressed and packaged video segments to the intelligent police service cloud platform, with the transmission priority of high-quality images being higher than that of compressed and packaged video segments.

[0107] S4, during transmission, the images of license plates and portraits extracted in S2 are transmitted first, and then the video segments and high-quality images are sorted according to the signal strength.

[0108] S4, a transmission queue is established, and the video segments, high-quality images, etc. to be transmitted are placed in the queue in order of transmission priority. Whenever new video data needs to be transmitted, the queue order is adjusted according to the current network signal strength and priority rules.

[0109] S5, after the intelligent police service cloud platform receives the license plate images and portrait images and video segments, it identifies the identity information, and then feeds back the identified identity information to the corresponding user of the law enforcement recorder according to the sending source, the specific steps are as follows:

[0110] Data reception: the intelligent police service cloud platform configures a special data reception service to listen to a specific network port and wait for the license plate images, portrait images and video segments uploaded by the law enforcement recorder;

[0111] Identity information identification: using license plate recognition algorithm to process the received license plate images, based on image processing technology and machine learning model, the characters and numbers on the license plate can be identified, and the identified license plate information is compared with the vehicle management database to obtain the related information of the vehicle, such as the name of the vehicle owner, the model of the vehicle, etc.

[0112] Face recognition technology is used to analyze the portrait images. By extracting the feature vector of the portrait, and comparing it with the known face database, the identity information of the person is determined;

[0113] Information feedback: collate the recognized identity information, including the vehicle owner information corresponding to the license plate and the character identity information corresponding to the portrait, associate the information with the identification of the sending source (law enforcement recorder) to accurately feedback to the corresponding user, so that the law enforcement personnel can timely understand the relevant situation.

[0114] The second purpose of the present application is to provide a video transmission system for a law enforcement recorder based on an intelligent police cloud platform, which comprises a video transmission method for a law enforcement recorder based on an intelligent police cloud platform according to any one of the above, and comprises a video acquisition unit 10, a video segment segmentation unit 20 and a feedback sending unit 30.

[0115] The video acquisition unit 10 is used for acquiring video data shot by the law enforcement recorder, and detecting the network signal strength of the connection between the law enforcement recorder and the intelligent police cloud platform.

[0116] The video segment segmentation unit 20 is used for establishing a brake light recognition model, then dividing the vehicles in the video data into to-be-detected vehicles and running vehicles according to the brake light recognition model, and simultaneously performing image extraction of license plates and portraits, then performing detection start marking and detection end marking, so as to complete video segment segmentation.

[0117] The feedback sending unit 30 is used for performing priority transmission allocation according to the network signal strength, and then the intelligent police cloud platform feeds back the recognized identity information according to the received data.

[0118] The above shows and describes the basic principles, main features and advantages of the present application. It should be understood by those skilled in the art that the present application is not limited by the above examples, and the above examples and descriptions in the specification are only preferred examples of the present application and are not intended to limit the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A video transmission method for law enforcement recorders based on an intelligent policing cloud platform, characterized in that: It comprises the following steps: S1, collecting the video data shot by the law enforcement recorder, while detecting the network signal strength of the connection between the law enforcement recorder and the intelligent police service cloud platform; S2, establishing a brake light recognition model, then dividing the vehicles in the video data into to-be-detected vehicles and running vehicles according to the brake light recognition model, and extracting the license plate and portrait images of the closest to-be-detected vehicle; S3, when the portrait image is extracted, detecting the start mark in the video data according to the time node of extraction, while setting the end time threshold, detecting the end mark according to the extracted new portrait image and the time threshold, and then segmenting the video clip in the video data according to the time node of the detection start mark and the detection end mark; S4, according to the network signal strength, the video clip is cached, and the video clip and the image of the license plate image and the portrait image are compressed and packaged, and then the priority transmission allocation is performed; S5, the intelligent police service cloud platform receives the license plate image and the portrait image and the video clip, and then identifies the identity information, and feeds back the identified identity information to the user corresponding to the law enforcement recorder according to the sending source; The steps of S2 are as follows: S2.1, collecting vehicle data in the network, and extracting image data of the brake light according to the vehicle data, and then establishing a brake light recognition model according to the image data; S2.2, combining the video data shot by the law enforcement recorder with the brake light recognition model to analyze the vehicles, extracting the vehicles in the video data respectively, and then identifying the extracted vehicles through the brake light recognition model, when the brake light recognition model identifies that the brake light of the vehicle is on, the vehicle is defined as a to-be-detected vehicle, otherwise, when the brake light recognition model identifies that the brake light of the vehicle is not on, the vehicle is defined as a running vehicle; S2.3, according to the real-time video data, the distance of the to-be-detected vehicle is analyzed to determine the closest to-be-detected vehicle to the law enforcement recorder, and then the license plate and the driver of the to-be-detected vehicle are identified in the video data, and when the license plate and the driver of the to-be-detected vehicle are identified in the video data, the license plate image and the portrait image of the driver are extracted. 2.The video transmission method for law enforcement recorders based on the intelligent police cloud platform according to claim 1, characterized in that: S1 establishes a network data transmission connection between the law enforcement recorder and the intelligent police service cloud platform, and then the law enforcement recorder continuously supplies the video data shot by the camera to the intelligent police service cloud platform for collection. 3.The video transmission method for law enforcement recorders based on the intelligent police cloud platform according to claim 1, characterized in that: S1 divides the network signal strength into high, medium and low; High network signal strength represents that the network speed can meet the real-time transmission of high-quality video data; Medium network signal strength represents that the network speed can meet the real-time transmission of low-quality video data, but not high-quality video data; Low network signal strength represents that the network speed does not meet the real-time transmission of video data.

4. The video transmission method for law enforcement recorders based on the intelligent police cloud platform according to claim 1, characterized in that: When extracting images, if the same segment cannot display the license plate and the driver at the same time, the license plate image and the portrait image of the driver are extracted respectively, and if the license plate and the driver appear in the same segment, the segment is extracted as the image of the license plate and the driver; Meanwhile, when the image of the license plate and the driver appears in the same segment has been extracted, stop continuing to extract, and cover the extracted license plate image and the portrait image of the driver.

5. The video transmission method for law enforcement recorders based on the intelligent police cloud platform according to claim 1, characterized in that: The steps of S3 are as follows: S3.1, when the portrait image is extracted, detection start marks are marked in the video data according to the time nodes of the extracted images; S3.2, set an end time threshold, when the video data is marked with the detection start, the video data does not appear the portrait image corresponding to the detection start mark, combine the end time threshold to start calculation, when the end time threshold is exceeded, the portrait image corresponding to the detection start mark still does not appear, that is, the detection end mark is marked in the video data; S3.3, when a new portrait image is extracted, the detection end mark is marked according to the time nodes of the extracted images, and the new detection start is marked at the same time; S3.4, when S3.2 and S3.3 mark the detection end mark, the video segment is segmented in the video data according to the time nodes of the detection start mark and the detection end mark, and the video segment dedicated to the portrait image is obtained.

6. The video transmission method for law enforcement recorders based on the intelligent police cloud platform according to claim 1, characterized in that: The steps of S4 are as follows: S4.1, according to the network signal strength, the video segment is cached, when the network signal strength is high, the video segment is not saved, when the network signal strength is medium and the network signal strength is low, the video segment is cached and compressed and packaged; S4.2, use the excess network space to supplement the cached video segment to the intelligent police cloud platform, and delete after sending; S4.3, when the network signal strength is in the high signal value, the law enforcement recorder transmits the video segment to the intelligent police cloud platform in high quality in real time; When the network signal strength is in the medium signal value, the law enforcement recorder transmits the compressed and packaged video segment to the intelligent police cloud platform; When the network signal strength is in the low signal value, the law enforcement recorder extracts the video segment in high quality in timing image, then transmits the extracted high quality image combined with the compressed and packaged video segment to the intelligent police cloud platform, the transmission priority of the high quality image is higher than that of the compressed and packaged video segment.

7. The video transmission method for law enforcement recorders based on the intelligent police cloud platform according to claim 1, characterized in that: In the transmission process of S4, the images of the license plate and the portrait extracted by S2 are preferentially transmitted, and then the video segment and the high quality image are sorted according to the signal strength.

8. A video transmission system for a law enforcement recorder based on an intelligent police cloud platform, comprising the video transmission method for a law enforcement recorder based on an intelligent police cloud platform in any one of claims 1-7, characterized in that: It comprises a video acquisition unit (10), a video segment segmentation unit (20) and a feedback sending unit (30); The video acquisition unit (10) is used to acquire the video data shot by the law enforcement recorder, and detect the network signal strength of the connection between the law enforcement recorder and the intelligent police cloud platform; The video segment segmentation unit (20) is used to establish a brake light recognition model, then divide the vehicle into a to-be-detected vehicle and a running vehicle in the video data according to the brake light recognition model, and extract the images of the license plate and the portrait, then mark the detection start and the detection end, so as to complete the video segment segmentation; The feedback sending unit (30) is used for priority transmission allocation according to network signal strength, and then the intelligent police cloud platform identifies the identity information according to the received data feedback.

Images