Data processing method and system

Abstract

The invention provides a data processing method and system, and the system comprises a plurality of types of collection devices and a plurality of road side calculation units, and each road side calculation unit is connected with the plurality of types of collection devices. The method comprises the following steps: a road side computing unit receives multi-source data sent by various types of acquisition equipment; the road side calculation unit carries out fusion processing on the multi-source data to obtain fusion data; the road side calculation unit generates at least one initial region of interest according to the fusion data; and the roadside calculation unit generates a target region of interest according to the fusion data and the at least one initial region of interest, so as to improve the generation efficiency and accuracy of the target region of interest.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a data processing method and system. Background Art

[0002] In the application scenarios of Internet of Vehicles and intelligent transportation, in order to improve the efficiency and real-time performance of perception data transmission and detection, it is necessary to crop the images or point cloud data of traffic events or obstacles to generate a region of interest (ROI).

[0003] Currently, ROIs are typically generated by cropping using a fixed, preset mechanism. This makes it impossible to adjust the area based on the dynamic spread of emergencies, and ROI updates are time-consuming, making it difficult to meet the millisecond-level real-time response requirements of connected vehicle scenarios. Furthermore, at the detection level, ROIs are typically independently triggered based on object detection by cameras or speed detection by millimeter-wave radars, resulting in a high false positive rate.

[0004] Therefore, there are certain limitations when generating ROI in the existing technology. Summary of the Invention

[0005] The purpose of this application is to provide a solution to the above-mentioned deficiencies in the prior art to solve the practical problems in the prior art.

[0006] To achieve the above objectives, the technical solutions adopted in the embodiments of the present application are as follows:

[0007] In a first aspect, an embodiment of the present application provides a data processing method, which is applied to a data processing system. The data processing system includes: a plurality of collection devices of various types and a plurality of roadside computing units, each of the roadside computing units being connected to the plurality of collection devices; the method includes:

[0008] The roadside computing unit receives multi-source data sent by various types of collection devices;

[0009] The roadside computing unit performs fusion processing on the multi-source data to obtain fused data;

[0010] The roadside calculation unit generates at least one initial region of interest based on the fused data;

[0011] The roadside calculation unit generates a target region of interest based on the fused data and the at least one initial region of interest.

[0012] As an optional implementation, the roadside calculation unit generates at least one initial region of interest based on the fused data, including:

[0013] The roadside calculation unit determines the confidence level of the target area where each target object is located based on the fused data, and uses the confidence level of the target area where each target object is located as the confidence level of each candidate region of interest; wherein the target object is an object whose acceleration or deceleration is greater than a first threshold;

[0014] The roadside calculation unit generates the at least one initial region of interest according to the fused data and the confidence of each candidate region of interest.

[0015] As an optional implementation, the roadside calculation unit generates a target region of interest based on the fused data and the at least one initial region of interest, including:

[0016] The roadside computing unit determines whether a target object accumulation event occurs based on the fused data;

[0017] If so, the roadside calculation unit expands the boundary of the at least one initial region of interest according to the fused data to generate the target region of interest.

[0018] As an optional implementation, the roadside calculation unit expands the boundary of the at least one initial region of interest based on the fused data to generate the target region of interest, including:

[0019] The roadside calculation unit merges the at least one initial region of interest into a region of interest to be adjusted;

[0020] The roadside calculation unit determines a boundary expansion coefficient based on the fused data;

[0021] The roadside calculation unit expands or contracts the region of interest to be adjusted according to the boundary expansion coefficient to generate the target region of interest.

[0022] As an optional implementation, the data processing system further includes: an edge server; each of the roadside computing units is communicatively connected to the edge server; and the method further includes:

[0023] The roadside computing unit adds multiple priority identifiers to the video frame where the target area of ​​interest is located, obtains multiple coding blocks and sends them to the edge server, wherein the priority identifiers are used to indicate the target coding quality of each area in the video frame;

[0024] The edge server receives the load parameters sent by each roadside computing unit and generates a load status table according to each load parameter;

[0025] The edge server determines each target roadside computing unit according to the load status table and the priority identifier in each coding block, and sends the target coding block to each target roadside computing unit;

[0026] The target roadside calculation unit performs region-by-region coding processing on the target coding block according to the priority identifier of the target coding block.

[0027] As an optional implementation, the roadside calculation unit adds multiple priority identifiers to the video frame where the target area of ​​interest is located to obtain multiple coding blocks, including:

[0028] The roadside calculation unit divides the video frame where the target area of ​​interest is located into a core area, a transition area and a background area;

[0029] The roadside calculation unit adds a first priority identifier to the core area to obtain a first coding block;

[0030] The roadside calculation unit adds a second priority identifier to the transition area to obtain a second coding block;

[0031] The roadside calculation unit adds a third priority identifier to the background area to obtain a third coding block.

[0032] As an optional implementation, the edge server determines each target roadside computing unit according to the load status table and the priority identifier in each coding block, including:

[0033] If the edge server determines, based on the priority identifier in the coding block, that the coding block is the first coding block, determining a first target roadside computing unit based on the load status table, wherein a utilization rate of a graphics processor of the first target roadside computing unit is less than a second threshold, and a current encoding quantity of the first coding block by the first target roadside computing unit is less than or equal to a third threshold;

[0034] If the edge server determines, based on the priority identifier in the coding block, that the coding block is the second coding block, determining a second target roadside computing unit based on the load status table, wherein a utilization rate of a graphics processor of the second target roadside computing unit is less than a fourth threshold;

[0035] If the edge server determines that the coding block is the third coding block based on the priority identifier in the coding block, a third target roadside computing unit is determined based on the load status table, wherein the utilization rate of the graphics processor of the third target roadside computing unit is less than a fifth threshold.

[0036] As an optional implementation, the target roadside calculation unit performs region-by-region coding processing on the target coding block according to the priority identifier of the target coding block, including:

[0037] The target roadside calculation unit determines a target coding rate and a target coding strategy according to the load parameter and the priority identifier of the target coding block;

[0038] The target roadside calculation unit performs region-by-region coding processing on the target coding blocks according to the target coding rate and the target coding strategy.

[0039] As an optional implementation, the target roadside calculation unit determines the target coding rate and the target coding strategy according to the load parameter and the priority identifier of the target coding block, including:

[0040] The first target roadside calculation unit determines a target coding rate for the first coding block according to the load parameter of the first target roadside calculation unit, a first preset weight of the core area, a second preset weight of the transition area, and a third preset weight of the background area, and determines that a target coding strategy for the first coding block is lossless coding;

[0041] The second target roadside calculation unit determines a target coding rate for the second coding block according to the load parameter of the second target roadside calculation unit, the first preset weight of the core area, the second preset weight of the transition area, and the third preset weight of the background area, and determines that the target coding strategy for the second coding block is medium quality coding;

[0042] The third target roadside calculation unit determines the target coding rate of the third coding block based on the load parameter of the third target roadside calculation unit, the first preset weight of the core area, the second preset weight of the transition area, and the third preset weight of the background area, and determines that the target coding strategy for the third coding block is low bit rate coding, and the third target roadside calculation unit and the first target roadside calculation unit perform staggered coding.

[0043] In a second aspect, an embodiment of the present application provides a data processing system, comprising: a plurality of collection devices of various types and a plurality of roadside computing units, each of the roadside computing units being connected to the plurality of collection devices;

[0044] The data processing system is used to execute the steps of the data processing method described in the first aspect above.

[0045] The beneficial effects of this application are:

[0046] The present application provides a data processing method and system. In the data processing system, each roadside computing unit is respectively connected to multiple types of acquisition devices to receive multi-source data sent by the different types of acquisition devices, and performs temporal and spatial fusion processing on the multi-source data to obtain spatiotemporally aligned fused data. The roadside computing unit analyzes the spatiotemporally aligned fused data to obtain motion information and visual information, and performs multimodal perception based on the motion information and visual information to generate at least one initial region of interest. Based on the motion information reflected by the spatiotemporally aligned fused data, the boundary of the initial region of interest is dynamically adjusted to generate a target region of interest. The roadside computing unit improves the generation efficiency and accuracy of the target region of interest through multimodal perception, so that the target region of interest covers the moving area of ​​each target in a timely and accurate manner, facilitates continuous tracking of each target in the target region of interest, and improves the accuracy and reliability of target tracking. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without creative work.

[0048] Figure 1 A schematic diagram of the architecture of a data processing system provided in an embodiment of the present application;

[0049] Figure 2 Schematic diagram of the data processing method provided in this embodiment Figure 1 ;

[0050] Figure 3 Schematic diagram of the data processing method provided in this embodiment Figure 2 ;

[0051] Figure 4 Schematic diagram of the data processing method provided in this embodiment Figure 3 ;

[0052] Figure 5 Schematic diagram of the data processing method provided in this embodiment Figure 4 ;

[0053] Figure 6 Another schematic diagram of the architecture of the data processing system provided in an embodiment of the present application;

[0054] Figure 7 Schematic diagram of the data processing method provided in this embodiment Figure 5 ;

[0055] Figure 8 Schematic diagram of the data processing method provided in this embodiment Figure 6 ;

[0056] Figure 9 Schematic diagram of the data processing method provided in this embodiment Figure 7 ;

[0057] Figure 10 Schematic diagram of the data processing method provided in this embodiment Figure 8 ;

[0058] Figure 11 Schematic diagram of the data processing method provided in this embodiment Figure 9 . DETAILED DESCRIPTION

[0059] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the drawings in the present application only serve the purpose of illustration and description and are not used to limit the scope of protection of the present application. In addition, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate the operations implemented according to some embodiments of the present application. It should be understood that the operations of the flowcharts can be implemented out of sequence, and steps without logical context can be reversed or implemented simultaneously. In addition, those skilled in the art, under the guidance of the contents of this application, can add one or more other operations to the flowchart, or remove one or more operations from the flowchart.

[0060] In addition, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application generally described and shown in the drawings here can be arranged and designed in various configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present application.

[0061] It should be noted that the term "comprising" will be used in the embodiments of the present application to indicate the existence of the features declared thereafter, but does not exclude the addition of other features.

[0062] In connected vehicles and intelligent transportation applications, ROIs are typically generated using a fixed, preset mechanism for cropping. This makes it impossible to adjust the area based on the dynamic spread of emergencies, and ROI updates are time-consuming, making it difficult to meet the millisecond-level real-time response requirements of connected vehicles. Furthermore, ROIs are typically independently triggered based on object detection by cameras or speed detection by millimeter-wave radars, resulting in a high false positive rate. This means that existing ROI generation techniques have certain limitations.

[0063] Based on the above-mentioned problems, an embodiment of the present application proposes a data processing method, which fuses multi-source data sent by various types of acquisition devices to obtain fused data, generates at least one initial region of interest based on the fused data and promptly captures target object accumulation events, and expands the boundaries of at least one initial region of interest when a target object accumulation event occurs to generate a target region of interest, ensuring that the target region of interest can promptly and accurately cover the development area of ​​the target object accumulation event.

[0064] Figure 1 The schematic diagram of the data processing system provided in the embodiment of the present application is as follows: Figure 1 As shown, the data processing system includes: multiple types of collection devices and multiple roadside computing units, and each roadside computing unit is connected to the multiple types of collection devices.

[0065] Optionally, the data processing system includes multiple types of acquisition devices, wherein the number of each type of acquisition device can be one or more. The multiple types can include, for example, types for acquiring images and types for acquiring point clouds. Each roadside computing unit is connected to multiple types of acquisition devices. For example, referring to Figure 1 The data processing system includes three roadside computing units (RCUs), each connected to an image acquisition device and a point cloud acquisition device. The various acquisition devices send their collected data to the RCUs, generating multi-source data. The RCUs are responsible for fusing the multi-source data and generating target ROIs.

[0066] Figure 2 Schematic diagram of the data processing method provided in this embodiment Figure 1 , the execution subject of this method is Figure 1 The data processing system shown in FIG. Figure 2 As shown, the method includes:

[0067] S101. A roadside computing unit receives multi-source data sent by various types of collection devices.

[0068] Optionally, the multiple types of acquisition devices may include image acquisition devices and point cloud acquisition devices. A single roadside computing unit is connected to both the image acquisition device and the point cloud acquisition device. The image sequence generated by the multi-source data image acquisition device and the point cloud data generated by the point cloud acquisition device are transmitted to the roadside computing unit, forming multi-source data. For example, the image acquisition device may be a camera, and the point cloud acquisition device may be a radar.

[0069] Specifically, the image acquisition device captures a sequence of images of the monitored area and transmits them in real time to a roadside computing unit connected to the image acquisition device. The point cloud acquisition device collects point cloud data of the monitored area and transmits the collected point cloud data at each moment in real time to a roadside computing unit connected to the point cloud acquisition device. The roadside computing unit receives the image sequence and point cloud data sent by the image acquisition device in real time. The point cloud data contains three-dimensional geometric information, while the image sequence contains two-dimensional color, shape, and texture information.

[0070] S102: The roadside computing unit performs fusion processing on the multi-source data to obtain fused data.

[0071] Optionally, the roadside computing unit fuses the image sequence and the point cloud data to obtain fused data. The roadside computing unit sequentially performs spatial alignment and temporal alignment on the two-dimensional image sequence and the three-dimensional point cloud data to ensure temporal and spatial consistency of the fused data.

[0072] Specifically, during the data fusion process, the roadside computing unit uses a calibration matrix to transform the polar coordinate system of the point cloud data and the pixel coordinate system of the image sequence, achieving spatial alignment between the image sequence and the point cloud data. The transformed point cloud data and image sequence are then time-synchronized to ensure that the timestamp deviation between the point cloud data and the image sequence is less than a preset duration, achieving temporal alignment between the image sequence and the point cloud data to produce fused data. For example, the preset duration can be 10 milliseconds.

[0073] S103: The roadside calculation unit generates at least one initial region of interest based on the fused data.

[0074] Optionally, the fused data includes motion information provided by the point cloud acquisition device and visual information provided by the image acquisition device. The roadside computing unit performs confidence-weighted fusion of multiple types of acquisition devices based on the motion information and visual information to generate at least one initial ROI.

[0075] Among them, multimodal perception is performed by combining visual information and motion information, so that the generated initial ROI can more accurately include each target and the dynamic area of ​​each target, thereby generating an initial ROI that is more in line with the actual situation and can effectively track each dynamic target.

[0076] S104: The roadside calculation unit generates a target region of interest based on the fused data and at least one initial region of interest.

[0077] Optionally, the roadside computing unit dynamically adjusts the boundary of the initial ROI according to the motion information reflected by the fused data to generate a target ROI.

[0078] Specifically, the roadside computing unit can dynamically adjust the boundaries of the initial ROI based on motion information through an exponential expansion algorithm, so that the generated target ROI can timely and accurately cover the moving area of ​​each target, thereby continuously tracking each target and improving the accuracy and reliability of target tracking.

[0079] In this embodiment, in the data processing system, each roadside computing unit is connected to various types of acquisition devices to receive multi-source data from these different types of acquisition devices. The unit then performs temporal and spatial fusion processing on the multi-source data to generate spatiotemporally aligned fused data. The roadside computing unit analyzes the spatiotemporally aligned fused data to obtain motion and visual information, and performs multimodal perception based on this motion and visual information to generate at least one initial region of interest (ROI). Based on the motion information reflected in the spatiotemporally aligned fused data, the unit dynamically adjusts the boundaries of the initial ROI to generate a target ROI. Through multimodal perception, the roadside computing unit improves the efficiency and accuracy of generating the target ROI, ensuring that the target ROI covers the moving area of ​​each target in a timely and accurate manner. This facilitates continuous tracking of each target within the target ROI and improves the accuracy and reliability of target tracking.

[0080] The following describes in detail the process of generating at least one initial region of interest by the roadside calculation unit based on the fused data.

[0081] Figure 3 Schematic diagram of the data processing method provided in this embodiment Figure 2 ,like Figure 3 As shown, in the above step S103, the roadside calculation unit generates at least one initial region of interest based on the fused data, including:

[0082] S201. A roadside computing unit determines, based on fused data, the confidence level of a target area where each target object is located, and uses the confidence level of the target area where each target object is located as the confidence level of each candidate region of interest; wherein the target object is an object having an acceleration or deceleration greater than a first threshold.

[0083] Optionally, the roadside computing unit determines each target object based on the motion information provided by the point cloud acquisition device in the fusion data, and regards the object with a sudden change in acceleration or deceleration greater than a first threshold as the target object. The image acquisition device is triggered to perform target dynamic detection on the area where each target object is located to obtain the target area where each target object is located. The motion information can be the motion speed. For example, the target dynamic detection algorithm can be the YOLOv7 algorithm, and the first threshold can be 10m / s 2 .

[0084] The roadside computing unit takes the target area where each target object is located detected by the image acquisition device as each candidate ROI, and performs weighted fusion of the confidence based on the motion information provided by the point cloud acquisition device and the visual information provided by the image acquisition device in the fusion data to obtain the confidence of the target area where each target object is located, that is, the confidence C of each candidate ROI is obtained. ROI .

[0085] Specifically, the confidence C of each candidate ROI is calculated based on the following formula: ROI :

[0086]

[0087] Among them, C ROI is the confidence of each candidate ROI, S is the visual detection score, which is used to characterize the confidence of the dynamic detection algorithm of the image acquisition device, γ is the weight of the visual detection score S, which can be 0.6, ΔV is the speed mutation of the target object, V TH is the speed mutation threshold, δ is the weight of the speed mutation, which can be 0.4.

[0088] Based on the above formula, in the confidence weighted fusion process, the velocity mutation amount ΔV of the target object and the velocity mutation amount threshold V TH The confidence C of the candidate ROI can be dynamically adjusted by taking the larger ratio of the target object's speed mutation amount and 1. ROI Specifically, when the target object's speed mutation ΔV is greater than the speed mutation threshold V TH When the target object's speed mutation ΔV and the speed mutation threshold V TH The ratio is greater than 1. At this time, in the confidence calculation, the target object's velocity mutation ΔV has a significant impact on the confidence C of the candidate ROI. ROI On the contrary, if the speed mutation amount ΔV of the target object is less than or equal to the speed mutation amount threshold V TH When the target object's speed mutation ΔV and the speed mutation threshold V TH If the ratio is less than or equal to 1, then 1 is taken to participate in the confidence calculation, indicating that the speed mutation amount ΔV of the target object at this time has a confidence C on the candidate ROI.ROI The impact of the speed mutation threshold V TH It can be set according to the actual application scenario to achieve the optimal confidence judgment effect.

[0089] S202: The roadside calculation unit generates at least one initial region of interest based on the fused data and the confidence level of each candidate region of interest.

[0090] Optionally, the roadside calculation unit calculates the confidence C of each candidate ROI according to ROI Determine the confidence of the target candidate ROI and convert the confidence C of each candidate ROI into ROI The confidence C of the candidate ROI that is greater than or equal to the preset confidence ROI The confidence level of the target candidate ROI is used as the confidence level. The roadside computing unit determines the visual detection area and motion speed of the target object corresponding to the confidence level of the target candidate ROI based on the fusion data, and generates at least one initial ROI. The preset confidence level can be 0.8. The size of the initial ROI is calculated based on the following formula: ROI :

[0091] W ROI =α×A obj +β×V obj

[0092] Among them, W ROI is the size of the initial ROI, A obj is the visual detection area of ​​the target object, V obj is the target object’s speed, α is the visual weight, and β is the motion weight. The visual weight α and motion weight β are used to characterize the effect of the target object’s visual detection area and motion speed on the size of the initial ROI W. ROI For example, α may be 0.5 and β may be 0.2.

[0093] In this embodiment, the roadside computing unit performs a weighted fusion of confidences based on the motion information provided by the point cloud acquisition device and the visual information provided by the image acquisition device in the fused data. This derives the confidence of the target area containing each target object with an acceleration or deceleration greater than a first threshold, and uses this confidence as the confidence of each candidate ROI. The roadside computing unit then determines the confidence of the target candidate ROI based on the confidence of each candidate ROI. Based on the fused data, the unit determines the visual detection area and motion speed of the target object corresponding to the confidence of the target candidate ROI, and generates at least one initial ROI. This initial ROI, generated by combining the visual detection area and motion speed of the target object, facilitates accurate dynamic tracking of the target object.

[0094] The following describes in detail a process in which the roadside calculation unit generates a target ROI based on the fused data and at least one initial ROI.

[0095] Figure 4 Schematic diagram of the data processing method provided in this embodiment Figure 3 ,like Figure 4 As shown, in the above step S104, the roadside computing unit generates a target region of interest based on the fused data and at least one initial region of interest, including:

[0096] S301. A roadside calculation unit determines whether a target object accumulation event occurs based on fused data.

[0097] Optionally, the roadside computing unit determines a point cloud density change rate in at least one initial ROI based on the fused data, and determines whether a target object accumulation event occurs based on the point cloud density change rate in at least one initial ROI, thereby improving event detection efficiency.

[0098] Specifically, if the growth rate of the point cloud density of consecutive frames in at least one initial ROI is greater than a preset ratio, the roadside computing unit determines that a target object accumulation event has occurred, for example, a target object accumulation event caused by sudden congestion or a traffic incident.

[0099] For example, if the growth rate of the point cloud density of five consecutive frames in at least one initial ROI is greater than 50%, the roadside computing unit may determine that a target object accumulation event has occurred.

[0100] S302: If yes, the roadside computing unit expands the boundary of at least one initial region of interest according to the fused data to generate a target region of interest.

[0101] Optionally, if the roadside computing unit determines that a target object accumulation event has occurred, it triggers the expansion of the boundary of at least one initial ROI to generate a target ROI.

[0102] Specifically, the roadside computing unit dynamically expands the boundaries of at least one initial ROI through an exponential expansion algorithm based on the motion information provided by the point cloud acquisition device in the fused data, and obtains a target ROI after boundary expansion, so as to generate a target ROI that is more consistent with the actual situation of the target object accumulation event, ensuring that the target ROI always covers the key information of the target object accumulation event.

[0103] In this embodiment, the roadside computing unit determines the rate of change in point cloud density within at least one initial region of interest (ROI) based on the fused data. Based on this rate of change, the unit determines whether a target object accumulation event has occurred. If so, the unit dynamically expands the boundaries of the at least one initial ROI using an exponential expansion algorithm based on the motion information provided by the point cloud acquisition device in the fused data, generating a target ROI with expanded boundaries. This improves the efficiency of detecting target object accumulation events and ensures that the generated target ROI consistently covers key information related to target object accumulation events.

[0104] The following describes in detail a process in which the roadside computing unit expands the boundary of at least one initial region of interest based on the fused data to generate a target region of interest.

[0105] Figure 5 Schematic diagram of the data processing method provided in this embodiment Figure 4 ,like Figure 5 As shown, in the above step S302, the roadside computing unit expands the boundary of at least one initial region of interest based on the fused data to generate a target region of interest, including:

[0106] S401: A roadside calculation unit merges at least one initial region of interest into a region of interest to be adjusted.

[0107] Optionally, after detecting a target object accumulation event, the roadside computing unit uses a bounding box merging algorithm to merge at least one initial ROI into an adjusted ROI that covers all target objects, ensuring that all target objects related to the target object accumulation event are within the range of the adjusted ROI.

[0108] Exemplarily, the initial ROIs are merged based on the coordinates, and the maximum boundary range of the ROI boxes of all target objects is found to form an ROI to be adjusted that can cover all targets.

[0109] S402: The roadside calculation unit determines a boundary expansion coefficient based on the fused data.

[0110] Optionally, the roadside computing unit determines the duration t and diffusion speed of the target object accumulation event based on the fusion data, and determines the boundary expansion coefficient based on the duration t and diffusion speed of the target object accumulation event, so as to dynamically adjust the boundary of the ROI to be adjusted based on the boundary expansion coefficient.

[0111] Specifically, the boundary expansion coefficient x is determined based on the following formula:

[0112]

[0113] Where x is the boundary expansion coefficient, e is the base, t is the duration of the target object accumulation event, V ave is the average velocity of the target object related to the target object accumulation event, V l The speed limit value is obtained.

[0114] S403: The roadside calculation unit expands or contracts the region of interest to be adjusted according to the boundary expansion coefficient to generate a target region of interest.

[0115] Optionally, the roadside calculation unit expands or contracts the ROI to be adjusted according to the boundary expansion coefficient x, and uses an exponential expansion algorithm to control the expansion or contraction amplitude of the ROI boundary to decay with the duration t of the target object accumulation event through an exponential term to generate a target ROI.

[0116] Specifically, the size B of the target ROI is determined based on the following formula: new :

[0117] B new =B o ×x

[0118] Among them, B new is the size of the target ROI, B o is the size of the ROI to be adjusted, and x is the boundary expansion coefficient.

[0119] That is, the size B of the target ROI is determined based on the following formula: new :

[0120]

[0121] Among them, B new is the size of the target ROI, B o is the size of the ROI to be adjusted, e is the base number, t is the duration of the target object accumulation event, V ave is the average velocity of the target object related to the target object accumulation event, V l The speed limit value is obtained.

[0122] In this embodiment, after the roadside computing unit detects the occurrence of a target object accumulation event, it uses a bounding box merging algorithm to merge at least one initial region of interest into a region of interest to be adjusted. Ensure that all target objects related to the target object accumulation event are within the region of interest to be adjusted. And based on the fused data, determine the duration and diffusion speed of the target object accumulation event, and determine the boundary expansion coefficient based on the duration and diffusion speed of the target object accumulation event. The roadside computing unit expands or contracts the region of interest to be adjusted according to the boundary expansion coefficient, and uses an exponential expansion algorithm to control the expansion or contraction amplitude of the boundary of the region of interest to be adjusted through an exponential term to decay with the duration of the target object accumulation event, thereby generating a target region of interest. The target region of interest is dynamically adjusted based on the duration and diffusion speed of the target object accumulation event to improve the update efficiency and accuracy of the target region of interest.

[0123] Figure 6 Another schematic diagram of the data processing system provided in the embodiment of the present application is shown in FIG. Figure 6 As shown, the data processing system also includes: an edge server; each roadside computing unit is communicatively connected to the edge server.

[0124] Optionally, refer to Figure 6 The data processing system also includes an edge server, with each roadside computing unit communicating with it. The edge server dispatches encoding tasks to the roadside computing units based on the load status and encoding blocks provided by the units, ensuring optimal allocation of network resources and load balancing. Each roadside computing unit encodes the video frame containing the target ROI in a region-by-region manner, based on the encoding tasks dynamically assigned by the edge server.

[0125] Figure 7 Schematic diagram of the data processing method provided in this embodiment Figure 5 ,like Figure 7 As shown, the method further includes:

[0126] S501. A roadside computing unit adds multiple priority identifiers to a video frame where a target region of interest is located, obtains multiple coding blocks, and sends them to an edge server. The priority identifiers are used to indicate target coding quality for each region in the video frame.

[0127] Optionally, the roadside computing unit dynamically adjusts the size B of the target ROI. new The video frame where the target ROI is located is divided into multiple regions, and multiple priority identifiers are added to the multiple regions in the video frame where the target ROI is located, forming multiple coding blocks and sending them to the edge server. The target coding quality of each region in the video frame is indicated by the priority identifier.

[0128] The coding block includes video data (ie, image sequence) and metadata. The metadata is used to describe the video data contained in the region, including the timestamp, attribute information, and priority tag of the video data in the region.

[0129] S502: The edge server receives the load parameters sent by each roadside computing unit, and generates a load status table according to each load parameter.

[0130] Optionally, the edge server communicates with each roadside computing unit at a preset time interval, sending a status query command to each roadside computing unit. Each roadside computing unit receives and responds to the status query command, sending its own real-time load parameters to the edge server. The load parameters include the roadside computing unit's central processing unit (CPU) utilization, graphics processing unit (GPU) utilization, memory usage, and network bandwidth usage. For example, the preset time interval can be 30 seconds.

[0131] The edge server records the real-time load parameters sent by each roadside computing unit and generates a load status table. The load status table includes the identifier of each roadside computing unit and the real-time load parameters. The edge server also updates the load status table based on the load parameters received from each roadside computing unit on a regular basis.

[0132] S503: The edge server determines each target roadside computing unit according to the load status table and the priority identifier in each coding block, and sends the target coding block to each target roadside computing unit.

[0133] Optionally, the edge server includes a software defined network (Software Defined Network) controller module ( Figure 6 (not shown), the SDN controller module in the edge server determines the target roadside computing units adapted to each coding block based on the load status table generated by the edge server and the priority identifier in each coding block sent by each roadside computing unit, and sends each target coding block adapted to each target roadside computing unit to each target roadside computing unit respectively, thereby realizing dynamic allocation of coding tasks.

[0134] Among them, the SDN controller module in the edge server can globally control the network topology and the load parameters of each roadside computing unit, perceive the computing power of each roadside computing unit in real time, and dynamically allocate corresponding encoding tasks to each target roadside computing unit based on the priority identifier in each encoding block.

[0135] S504. The target roadside calculation unit performs region-by-region coding processing on the target coding block according to the priority identifier of the target coding block.

[0136] Optionally, the target roadside calculation unit performs region-by-region encoding processing on the target coding block according to the target coding quality indicated by the priority identifier of the target coding block, and embeds user experience quality (Quality of Experience, referred to as QoE) metadata into the encoded video data, wherein the QoE metadata includes quality weight information of each region and target parameters related to Peak Signal-to-Noise Ratio (PSNR), so as to better manage and optimize user experience during data transmission and processing.

[0137] Each target roadside computing unit can rationally allocate limited network and storage resources to target coding blocks of different priorities. Specifically, higher-priority coding blocks are allocated more resources to ensure their accuracy and efficiency, while lower-priority coding blocks can be appropriately reduced in resource usage to improve resource utilization without affecting overall performance.

[0138] In this embodiment, the roadside computing unit divides the video frame containing the target ROI into multiple regions based on the size of the target ROI. It then adds multiple priority identifiers to each of the regions within the video frame, creating multiple encoding blocks that are sent to the edge server. The priority identifiers indicate the target encoding quality for each region within the video frame. The edge server receives the load parameters sent by each roadside computing unit and generates a load status table. Based on the load status table and the priority identifiers in each encoding block, it determines the target roadside computing unit and issues the target encoding block to each target roadside computing unit. The target roadside computing unit then performs region-by-region encoding on the target encoding block based on the target encoding quality indicated by the priority identifier of the target encoding block. By real-time sensing of the computing power of each roadside computing unit and combining the priority identifiers in each encoding block, the target roadside computing unit dynamically allocates corresponding encoding tasks to the target roadside computing unit, allowing each target roadside computing unit to allocate network and storage resources to target encoding blocks of different priorities. This improves resource allocation rationality and resource utilization.

[0139] The following describes in detail the process in which the roadside computing unit adds multiple priority identifiers to the video frame where the target region of interest is located to obtain multiple coding blocks.

[0140] Figure 8 Schematic diagram of the data processing method provided in this embodiment Figure 6 ,like Figure 8 As shown, in the above step S501, the roadside calculation unit adds multiple priority identifiers to the video frame where the target area of ​​interest is located to obtain multiple coding blocks, including:

[0141] S601: A roadside calculation unit divides a video frame where a target area of ​​interest is located into a core area, a transition area, and a background area.

[0142] Optionally, the roadside calculation unit is based on the size B of the target ROI new The video frame where the target ROI is located is divided into a core region, a transition region, and a background region based on different distance ranges from the center of the target ROI. The core region is the area within a first preset range from the center of the target ROI, the transition region is the area within a second preset range outside the core region, and the background region is the area outside the core region and transition region. The core region contains the largest amount of data and is therefore of the highest importance. The background region contains the smallest amount of data and is therefore of the lowest importance.

[0143] For example, the area within a radius of 10m with the center of the target ROI as the origin is divided into the core area, the area within a 20m range outside the core area is divided into the transition area, and the rest of the area in the video frame except the core area and the transition area is divided into the background area.

[0144] S602: The roadside calculation unit adds a first priority identifier to the core area to obtain a first coding block.

[0145] Optionally, the roadside computing unit adds a first priority identifier to the core area, obtains metadata of the core area, and forms a first coding block based on the video data and metadata of the core area. The first priority identifier is used to indicate a target coding quality of the video data of the core area.

[0146] The metadata of the core area is structured data used to describe the video data of the core area, including a timestamp, attribute information, and a first priority tag of the video data of the core area. For example, the first priority tag may be 0xFFFF.

[0147] S603: The roadside calculation unit adds a second priority identifier to the transition area to obtain a second coding block.

[0148] Optionally, the roadside computing unit adds a second priority identifier to the transition area, obtains metadata of the transition area, and forms a second coding block based on the video data and metadata of the transition area. The second priority identifier is used to indicate a target coding quality of the video data of the transition area.

[0149] The metadata of the transition region is structured data used to describe the video data of the transition region, including a timestamp, attribute information, and a second priority tag of the video data of the transition region. For example, the second priority tag may be 0x7FFF.

[0150] S604: The roadside calculation unit adds a third priority identifier to the background area to obtain a third coding block.

[0151] Optionally, the roadside calculation unit adds a third priority identifier to the background area to obtain metadata of the background area, and forms a third coding block based on the video data and metadata of the background area. The third priority identifier is used to indicate a target coding quality of the video data of the background area.

[0152] The metadata of the background area is structured data used to describe the video data of the background area, including a timestamp, attribute information, and a third priority tag of the video data of the background area. For example, the third priority tag may be 0x0001.

[0153] In this embodiment, the roadside computing unit divides the video frame containing the target ROI into a core region, a transition region, and a background region based on the size of the target ROI. It then adds a first priority identifier, a second priority identifier, and a third priority identifier to the core region, the transition region, and the background region, respectively, to form a first coding block, a second coding block, and a third coding block. Each priority identifier indicates the target coding quality of the video data for each region, allowing coding to be performed region by region according to the corresponding priority identifier in each coding block, thereby conserving network bandwidth resources.

[0154] The following describes in detail the process of the edge server determining each target roadside computing unit according to the load status table and the priority identifiers in each coding block.

[0155] Figure 9 Schematic diagram of the data processing method provided in this embodiment Figure 7 ,like Figure 9 As shown, in the above step S503, the edge server determines each target roadside computing unit according to the load status table and the priority identifier in each coding block, including:

[0156] S701. If the edge server determines that the coding block is the first coding block based on the priority identifier in the coding block, the first target roadside computing unit is determined based on the load status table, wherein the utilization rate of the graphics processor of the first target roadside computing unit is less than the second threshold, and the current number of codes for the first coding block by the first target roadside computing unit is less than or equal to the third threshold.

[0157] Optionally, if the SDN controller module in the edge server determines that the priority identifier in the coding block is the first priority identifier, that is, determines that the coding block is the first coding block, then the load status table is searched for a first target roadside computing unit that meets a first preset condition. The first preset condition is that the GPU utilization rate of the roadside computing unit is less than a second threshold, and the number of codes currently being encoded for the first coding block in the roadside computing unit is less than or equal to a third threshold.

[0158] That is, the SDN controller in the edge server uses the roadside computing unit whose GPU utilization is less than the second threshold and whose current encoding quantity for the first coding block is less than or equal to the third threshold as the target roadside computing unit.

[0159] For example, the second threshold may be 70%, and the third threshold may be 5. That is, the GPU utilization of the first target roadside computing unit is less than 70%, and the current number of codes for the first coding block is less than or equal to 5.

[0160] S702: If the edge server determines that the coding block is the second coding block based on the priority identifier in the coding block, a second target roadside computing unit is determined based on the load status table, wherein the utilization rate of the graphics processor of the second target roadside computing unit is less than a fourth threshold.

[0161] Optionally, if the SDN controller module in the edge server determines that the priority identifier in the coding block is the second priority identifier, that is, determines that the coding block is the second coding block, then the load status table is queried for a second target roadside computing unit that meets a second preset condition. The second preset condition is that the GPU utilization of the roadside computing unit is less than a fourth threshold.

[0162] That is, the SDN controller in the edge server uses the roadside computing unit whose GPU utilization is less than the fourth threshold as the target roadside computing unit.

[0163] For example, the fourth threshold may be 50%, that is, the GPU utilization rate of the second target roadside computing unit is less than 50%.

[0164] S703. If the edge server determines that the coding block is the third coding block based on the priority identifier in the coding block, a third target roadside computing unit is determined based on the load status table, wherein the utilization rate of the graphics processor of the third target roadside computing unit is less than the fifth threshold.

[0165] Optionally, if the SDN controller module in the edge server determines that the priority identifier in the coding block is the third priority identifier, that is, determines that the coding block is the third coding block, then a third target roadside computing unit that meets a third preset condition is queried from the load status table. The third preset condition is that the GPU utilization of the roadside computing unit is less than a fifth threshold.

[0166] That is, the SDN controller in the edge server uses the roadside computing unit whose GPU utilization is less than the fifth threshold as the target roadside computing unit.

[0167] For example, the fifth threshold may be 30%, that is, the GPU utilization rate of the third target roadside computing unit is less than 30%.

[0168] In this embodiment, if the edge server determines that the coding block is the first coding block based on the priority identifier in the coding block, the first target roadside computing unit adapted to the first coding block is determined according to the load status table. If the edge server determines that the coding block is the second coding block based on the priority identifier in the coding block, the second target roadside computing unit adapted to the second coding block is determined according to the load status table. If the edge server determines that the coding block is the third coding block based on the priority identifier in the coding block, the third target roadside computing unit adapted to the first coding block is determined according to the load status table. The first target roadside computing unit with higher performance is enabled to complete the encoding task of the first coding block with higher importance. The adaptation of each coding block to each target roadside computing unit is achieved, ensuring reasonable resource allocation when encoding coding blocks of each priority.

[0169] The following describes in detail the process in which the target roadside calculation unit performs region-by-region coding processing on the target coding block according to the priority identifier of the target coding block.

[0170] Figure 10 Schematic diagram of the data processing method provided in this embodiment Figure 8 ,like Figure 10 As shown, in the above step S504, the target roadside calculation unit performs region-by-region coding processing on the target coding block according to the priority identifier of the target coding block, including:

[0171] S801. The target roadside calculation unit determines a target coding rate and a target coding strategy according to a load parameter and a priority identifier of a target coding block.

[0172] Optionally, the target roadside calculation unit calculates the target coding rate of the target coding block according to the load parameter and the priority identifier of the target coding block, and determines the target coding strategy corresponding to the target coding block based on the priority identifier of the target coding block.

[0173] Among them, for high-priority target coding blocks, a higher coding rate and a higher-quality coding strategy will be adopted to ensure the accuracy and completeness of the information; while for low-priority target coding blocks, a lower coding rate and a lower-quality coding strategy will be adopted to reduce the occupation of computing resources while ensuring a certain information quality.

[0174] S802: The target roadside calculation unit performs region-by-region coding processing on the target coding blocks according to the target coding rate and the target coding strategy.

[0175] Optionally, after determining the target coding rate and target coding strategy, the target roadside calculation unit performs region-by-region coding on the target coding block according to the region to which the target coding block belongs, and embeds the QoE metadata corresponding to the target coding block into the encoded video data.

[0176] For the area to which the target coding block belongs, the target coding block is coded in different areas according to the target coding rate and target coding strategy of the target coding block, which can avoid the computing power overload of each roadside computing unit and reduce idle network resources.

[0177] After encoding each region, the target roadside computing unit can upload the encoded data of each region to the upper-level platform. Specifically, a dual-link redundant transmission strategy is used to transmit the encoded data of the core region, and the hash value is verified to ensure the integrity of the encoded data of the core region.

[0178] A transmission strategy based on dynamic bandwidth adjustment is adopted to transmit the coded data of the transition area. When the network bandwidth is sufficient, a high-quality single link is used to transmit the coded data of the transition area, and CRC code verification data is embedded. When the network bandwidth is medium, single-link lossless compression is adopted to transmit the coded data of the transition area and CRC code verification data is embedded. When the network bandwidth is tight, a strict rate-elastic compression algorithm (such as one based on wavelet transform) is adopted to transmit the code of the transition area.

[0179] The "on-demand pull" mode or code rate elastic compression strategy is used to transmit the encoded data of the background area. When the network bandwidth is tight, it switches to the black frame filling mode and only transmits the metadata in the encoded data of the background area, reducing unnecessary data transmission and improving bandwidth utilization.

[0180] In this embodiment, the target roadside computing unit determines the target coding rate and target coding strategy for the target coding block based on the load parameters and the priority identifier of the target coding block. Based on the target coding rate and target coding strategy, the target coding block is coded by region. This avoids computing power overload on each roadside computing unit, reduces idle network resources, and improves resource utilization.

[0181] The following describes in detail the process of determining the target coding rate and the target coding strategy by the target roadside calculation unit according to the load parameters and the priority identifiers of the target coding blocks.

[0182] Figure 11 Schematic diagram of the data processing method provided in this embodiment Figure 9 ,like Figure 11 As shown, in the above step S801, the target roadside calculation unit determines the target coding rate and the target coding strategy according to the load parameter and the priority identifier of the target coding block, including:

[0183] S901. The first target roadside calculation unit determines the target coding rate of the first coding block based on the load parameters of the first target roadside calculation unit, the first preset weight of the core area, the second preset weight of the transition area, and the third preset weight of the background area, and determines that the target coding strategy for the first coding block is lossless coding.

[0184] Optionally, the first target roadside calculation unit determines the target coding rate R of the first coding block based on the following formula according to the load parameter of the first target roadside calculation unit, the first preset weight of the core area, the second preset weight of the transition area, and the third preset weight of the background area. h :

[0185]

[0186] Among them, R h is the target coding rate of the first coding block, B avl1 is the network bandwidth occupancy status of the first target roadside computing unit, Q h is the quality weight of the core area to which the first coding block belongs, Q g is the quality weight of the transition region to which the second coding block belongs, Q b is the quality weight of the background area to which the third coding block belongs.

[0187] For example, Q h is 1, Q g is 0.6, Q b is 0.2.

[0188] Because the area to which the first coding block belongs is the core area and has relatively high coding quality requirements, the target coding strategy for the first coding block is determined to be lossless coding. Specifically, the first coding block in the core area adopts the H.266VVC lossless coding strategy, with a quantization parameter (QP) of 23 and a group of pictures (GOP) of 1, ensuring clear texture of the key target of the first coding block and a PSNR of ≥ 42dB.

[0189] Among them, QP is used to represent the degree of quantization fineness in the encoding process, and GOP is used to represent the grouping method of continuous frames in video data.

[0190] S902. The second target roadside calculation unit determines the target coding rate of the second coding block based on the load parameters of the second target roadside calculation unit, the first preset weight of the core area, the second preset weight of the transition area, and the third preset weight of the background area, and determines that the target coding strategy for the second coding block is medium quality coding.

[0191] Optionally, the second target roadside calculation unit determines the target coding rate R of the second coding block based on the following formula according to the load parameter of the second target roadside calculation unit, the first preset weight of the core area, the second preset weight of the transition area, and the third preset weight of the background area. g :

[0192]

[0193] Among them, R g is the target coding rate of the second coding block, B avl2 is the network bandwidth occupancy status of the second target roadside computing unit, Q h is the quality weight of the core area to which the first coding block belongs, Q g is the quality weight of the transition region to which the second coding block belongs, Q b is the quality weight of the background area to which the third coding block belongs.

[0194] Because the second coding block belongs to a transition region with encoding quality requirements between the core and background regions, the target encoding strategy for the second coding block is determined to be medium-quality encoding. Specifically, the second coding block in the transition region adopts the medium-quality encoding strategy of High Efficiency Video Coding (HEVC), with a QP of 28 and a GOP of 8, retaining only the outline information of the object.

[0195] S903. The third target roadside calculation unit determines the target coding rate of the third coding block according to the load parameters of the third target roadside calculation unit, the first preset weight of the core area, the second preset weight of the transition area, and the third preset weight of the background area, and determines that the target coding strategy for the third coding block is low bit rate coding, and the third target roadside calculation unit and the first target roadside calculation unit perform staggered coding.

[0196] Optionally, the third target roadside calculation unit determines the target coding rate R of the third coding block based on the following formula according to the load parameter of the third target roadside calculation unit, the first preset weight of the core area, the second preset weight of the transition area, and the third preset weight of the background area. b :

[0197]

[0198] Among them, R b is the target coding rate of the third coding block, B avl3 is the network bandwidth occupancy status of the second target roadside computing unit, Q h is the quality weight of the core area to which the first coding block belongs, Q g is the quality weight of the transition region to which the second coding block belongs, Q b is the quality weight of the background area to which the third coding block belongs.

[0199] Because the third coding block belongs to the background area, which has relatively low encoding quality requirements, the target encoding strategy for the third coding block is determined to be low bitrate encoding, and the third target roadside calculation unit and the first target roadside calculation unit perform encoding at staggered intervals. Specifically, the third coding block in the background area uses the AV1 low bitrate encoding strategy, with a QP of 45 and a GOP of 23 to reduce data volume.

[0200] Among them, the staggered coding of the third target roadside computing unit and the first target roadside computing unit is achieved by aggregating the third coding blocks of multiple background areas into batch coding tasks through the SDN controller module in the edge server, and assigning them to the third target roadside computing unit, and instructing the third target roadside computing unit to perform encoding processing of the third coding blocks during non-peak periods, that is, staggered coding with the first target roadside computing unit to achieve load balancing.

[0201] In this embodiment, each target roadside computing unit adaptively determines the target coding rate and corresponding target coding strategy for each coding block based on its load parameters and the preset weights for each region. The target coding strategy for the first coding block is lossless coding, the target coding strategy for the second coding block is medium-quality coding, and the target coding strategy for the third coding block is low-rate coding. The third target roadside computing unit performs staggered coding operations compared to the first target roadside computing unit. This achieves dynamic allocation of coding rates, ensuring that appropriate coding strategies and rates are assigned to coding blocks in different regions, effectively conserving bandwidth resources.

[0202] An embodiment of the present application further provides a data processing system, which includes: multiple collection devices of various types and multiple roadside computing units, each roadside computing unit being connected to the multiple types of collection devices.

[0203] The data processing system is used to execute the steps of the data processing method described in the above embodiments.

[0204] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system and device described above can refer to the corresponding process in the method embodiment, and will not be repeated in this application. In the several embodiments provided in this application, it should be understood that the disclosed system, device and method can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0205] In addition, the functional units in the various embodiments of the present application can be integrated into a processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in the various embodiments of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0206] The above is only a specific implementation method of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the protection scope of the present application.

Claims

1. A data processing method, characterized in that: Applied to a data processing system, the data processing system comprising: a plurality of collection devices of various types and a plurality of roadside computing units, each of the roadside computing units being connected to the plurality of collection devices; the method comprising: The roadside computing unit receives multi-source data sent by various types of collection devices; The roadside computing unit performs fusion processing on the multi-source data to obtain fused data; The roadside calculation unit generates at least one initial region of interest based on the fused data; The roadside calculation unit generates a target region of interest based on the fused data and the at least one initial region of interest.

2. The method according to claim 1, characterized in that The roadside calculation unit generates at least one initial region of interest based on the fused data, including: The roadside calculation unit determines the confidence level of the target area where each target object is located based on the fused data, and uses the confidence level of the target area where each target object is located as the confidence level of each candidate region of interest; wherein the target object is an object whose acceleration or deceleration is greater than a first threshold; The roadside calculation unit generates the at least one initial region of interest according to the fused data and the confidence of each candidate region of interest.

3. The method according to claim 1, characterized in that The roadside calculation unit generates a target region of interest based on the fused data and the at least one initial region of interest, including: The roadside computing unit determines whether a target object accumulation event occurs based on the fused data; If so, the roadside calculation unit expands the boundary of the at least one initial region of interest according to the fused data to generate the target region of interest.

4. The method according to claim 3, characterized in that The roadside calculation unit performs boundary expansion on the at least one initial region of interest based on the fused data to generate the target region of interest, including: The roadside calculation unit merges the at least one initial region of interest into a region of interest to be adjusted; The roadside calculation unit determines a boundary expansion coefficient based on the fused data; The roadside calculation unit expands or contracts the region of interest to be adjusted according to the boundary expansion coefficient to generate the target region of interest.

5. The method according to any one of claims 1 to 4, characterized in that The data processing system further includes: an edge server; each of the roadside computing units is in communication with the edge server; and the method further includes: The roadside computing unit adds multiple priority identifiers to the video frame where the target area of ​​interest is located, obtains multiple coding blocks and sends them to the edge server, wherein the priority identifiers are used to indicate the target coding quality of each area in the video frame; The edge server receives the load parameters sent by each roadside computing unit and generates a load status table according to each load parameter; The edge server determines each target roadside computing unit according to the load status table and the priority identifier in each coding block, and sends the target coding block to each target roadside computing unit; The target roadside calculation unit performs region-by-region coding processing on the target coding block according to the priority identifier of the target coding block.

6. The method according to claim 5, characterized in that The roadside calculation unit adds multiple priority identifiers to the video frame where the target area of ​​interest is located to obtain multiple coding blocks, including: The roadside calculation unit divides the video frame where the target area of ​​interest is located into a core area, a transition area and a background area; The roadside calculation unit adds a first priority identifier to the core area to obtain a first coding block; The roadside calculation unit adds a second priority identifier to the transition area to obtain a second coding block; The roadside calculation unit adds a third priority identifier to the background area to obtain a third coding block.

7. The method according to claim 6, characterized in that The edge server determines each target roadside computing unit according to the load status table and the priority identifier in each coding block, including: If the edge server determines, based on the priority identifier in the coding block, that the coding block is the first coding block, determining a first target roadside computing unit based on the load status table, wherein a utilization rate of a graphics processor of the first target roadside computing unit is less than a second threshold, and a current encoding quantity of the first coding block by the first target roadside computing unit is less than or equal to a third threshold; If the edge server determines, based on the priority identifier in the coding block, that the coding block is the second coding block, determining a second target roadside computing unit based on the load status table, wherein a utilization rate of a graphics processor of the second target roadside computing unit is less than a fourth threshold; If the edge server determines that the coding block is the third coding block based on the priority identifier in the coding block, a third target roadside computing unit is determined based on the load status table, wherein the utilization rate of the graphics processor of the third target roadside computing unit is less than a fifth threshold.

8. The method according to claim 5, characterized in that The target roadside calculation unit performs region-by-region coding processing on the target coding block according to the priority identifier of the target coding block, including: The target roadside calculation unit determines a target coding rate and a target coding strategy according to the load parameter and the priority identifier of the target coding block; The target roadside calculation unit performs region-by-region coding processing on the target coding blocks according to the target coding rate and the target coding strategy.

9. The method according to claim 6, characterized in that The target roadside calculation unit determines a target coding rate and a target coding strategy according to the load parameter and the priority identifier of the target coding block, including: The first target roadside calculation unit determines a target coding rate for the first coding block according to the load parameter of the first target roadside calculation unit, a first preset weight of the core area, a second preset weight of the transition area, and a third preset weight of the background area, and determines that a target coding strategy for the first coding block is lossless coding; The second target roadside calculation unit determines a target coding rate for the second coding block according to the load parameter of the second target roadside calculation unit, the first preset weight of the core area, the second preset weight of the transition area, and the third preset weight of the background area, and determines that the target coding strategy for the second coding block is medium quality coding; The third target roadside calculation unit determines the target coding rate of the third coding block based on the load parameter of the third target roadside calculation unit, the first preset weight of the core area, the second preset weight of the transition area, and the third preset weight of the background area, and determines that the target coding strategy for the third coding block is low bit rate coding, and the third target roadside calculation unit and the first target roadside calculation unit perform staggered coding.

10. A data processing system, characterized in that: The data processing system includes: a plurality of collection devices of various types and a plurality of roadside computing units, each of the roadside computing units being connected to the plurality of collection devices; The data processing system is used to execute the steps of the data processing method according to any one of claims 1 to 9.

Images