An information data collection method and system based on artificial intelligence

Abstract

The application is suitable for the technical field of information data collection, and particularly relates to an information data collection method and system based on artificial intelligence. The method comprises the following steps: building an information data collection platform, determining a registration device, selecting a collection unit and a processing unit, identifying a network service device in each processing unit, and building an edge network; using a video monitoring device integrated in the collection unit in advance to collect video data, intercepting a segment containing roadside features, generating an access request when the collection unit scans a service signal of the edge network, establishing a communication link between the collection unit and the network service device, and receiving a use permission issued by a user. The application can estimate roadside pedestrian flow by calculating pedestrian flow and using video data in the collection unit, greatly reduces infrastructure deployment and operation and maintenance costs, greatly widens collection sources, and effectively guarantees timeliness and accuracy of information data.

Description

Technical Field

[0001] The present invention relates to the technical field of information data collection, and in particular to an information data collection method and system based on artificial intelligence. Background Art

[0002] Information data collection using artificial intelligence refers to the process of using artificial intelligence technologies (such as computer vision, natural language processing, speech recognition, and deep learning) to automatically identify, analyze, and extract data from multiple sources such as video surveillance, sensors, social platforms, and text materials.

[0003] Vehicle flow can reflect the operating conditions of urban roads and the level of traffic load, while pedestrian flow data is of great significance in many fields such as urban management, commercial operations and public safety. In existing technologies, video surveillance equipment or infrared sensors are generally deployed on both sides of the street to collect pedestrian flow data in real time, helping managers understand the density and flow trends of pedestrians in different time periods and regions; however, this not only increases the initial construction and maintenance costs, but also puts high demands on data transmission, storage and privacy protection.

[0004] Therefore, “how to use the driving recorder and the idle network to estimate the crowd density” is the technical problem that the present invention needs to solve. Summary of the Invention

[0005] The purpose of the present invention is to provide an information data collection method and system based on artificial intelligence to solve the problem of "how to use driving recorder data and idle network to estimate crowd density" raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] An information data collection method based on artificial intelligence, the method comprising:

[0008] Build an information data collection platform, determine the registration equipment, select the collection unit and processing unit, identify the network service equipment in each processing unit, and build the edge network;

[0009] The video surveillance equipment pre-integrated in the acquisition unit is used to collect video data and extract segments containing roadside features. When the acquisition unit scans the service signal of the edge network, it generates an access request and establishes a communication link between the acquisition unit and the network service device.

[0010] receiving the usage permission issued by the user, and uploading the clip to the processing unit via the communication link;

[0011] Construct a pedestrian recognition model and deploy it to the processing unit. Input the clip into the pedestrian recognition model, output the number of people in each video frame, and calculate the flow of people in each clip by superposition. Obtain attribute data of each clip, wherein the attribute data includes at least: location and time;

[0012] Write the traffic and attribute data into the fragment, generate the collection results, and upload them to the information data collection platform.

[0013] Furthermore, the steps of determining the registration device, selecting the collection unit and the processing unit, identifying the network service device in each processing unit, and building the edge network include:

[0014] Configure the evaluation indicators of the processing unit, including at least: equipment utilization and transmission delay;

[0015] Based on the evaluation index, a guidance mechanism with the collection unit as the object is created.

[0016] Furthermore, the step of building an edge network includes:

[0017] Calculating the number of concurrent connections of each edge network, selecting a set value, and generating an avoidance reminder based on the attribute data when the number of concurrent connections is greater than the set value;

[0018] A coverage range is constructed according to the service signal, and when a collection unit enters the coverage range, the avoidance reminder is sent to the collection unit.

[0019] Furthermore, the step of using the video surveillance equipment pre-integrated in the acquisition unit to collect video data and extracting a segment containing roadside features includes:

[0020] Define the collection scope of information data and draw a planar road network map within the collection scope;

[0021] Traffic light intersections are marked, and several road sections are cut out from the planar road network map. A time window corresponding to each road section is created. When the collection results are received, the time window is shifted.

[0022] Furthermore, the steps of constructing a pedestrian recognition model, deploying it in a processing unit, and inputting the fragment into the pedestrian recognition model include:

[0023] Inserting a target detection module into the pedestrian recognition model, collecting historical video data, annotating it, generating a training set, and training the target detection module;

[0024] The segments are input into a target detection module in parallel, and multi-target features are output, wherein the multi-target features at least include: roadside vendors and roadside vehicles.

[0025] Furthermore, the method further comprises:

[0026] The segments where the multi-target features are located are defined as target segments, the target segments and attribute data are integrated to generate a roadside information set, and then uploaded to the preset platform;

[0027] Through the information data collection platform, the user's positioning permission is obtained, the location of the target segment is used as the center of the circle, the preset distance is used as the radius, the verification range is defined, the verification task is generated, and it is published to the information data collection platform.

[0028] Furthermore, the method further comprises:

[0029] Establish the corresponding relationship between the flow of people and the time window, and select the middle point from the time window;

[0030] Use the middle point as the horizontal coordinate and the corresponding flow of people as the vertical coordinate to draw a trend graph.

[0031] Furthermore, the system includes:

[0032] The construction module is used to build an information data collection platform, determine the registration device, select the collection unit and processing unit, identify the network service device in each processing unit, and build the edge network;

[0033] An upload module is configured to utilize a video surveillance device pre-integrated in the acquisition unit to collect video data and extract segments containing roadside features. When the acquisition unit scans a service signal from the edge network, it generates an access request, establishes a communication link between the acquisition unit and the network service device, receives a user authorization, and uploads the segments to the processing unit via the communication link.

[0034] A construction module is used to build a pedestrian recognition model, deploy it to the processing unit, input the fragment into the pedestrian recognition model, output the number of people in each video frame, and superimpose and calculate the flow of people in each fragment, obtain the attribute data of each fragment, where the attribute data at least includes: location and time, write the flow of people and attribute data into the fragment, generate the collection results, and upload them to the information data collection platform.

[0035] Furthermore, the building module includes:

[0036] A configuration unit, configured to configure evaluation indicators of the processing unit, including at least: device utilization and transmission delay;

[0037] A creation unit, configured to create a guidance mechanism with the collection unit as the object according to the evaluation index;

[0038] a calculation unit, configured to calculate the number of concurrent connections of each edge network, select a set value, and generate an avoidance reminder based on the attribute data when the number of concurrent connections is greater than the set value;

[0039] The sending unit is used to build a coverage range according to the service signal, and send the avoidance reminder to the collection unit when the collection unit enters the coverage range.

[0040] Furthermore, the upload module includes:

[0041] A delineation unit is used to delineate the collection range of information data and draw a planar road network map within the collection range;

[0042] The offset unit is used to mark the traffic light intersection and cut out several road sections from the planar road network map, create a time window corresponding to each road section, and offset the time window after receiving the collection result.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] By building an edge network, the timeliness and coverage of pedestrian flow data can be improved, while making full use of idle network resources to greatly improve the efficiency of information data collection. By building a communication link, video data can be sent to the edge network for processing, realizing distributed data processing and improving edge network utilization. By calculating pedestrian flow, the video data in the collection unit can be used to estimate roadside pedestrian flow, which greatly reduces the cost of infrastructure deployment and operation and maintenance, greatly broadens the data source, and effectively guarantees the timeliness and accuracy of information data. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 A flowchart of an information data collection method based on artificial intelligence provided by an embodiment of the present invention;

[0046] Figure 2 A block diagram of the first sub-process of the information data collection method based on artificial intelligence provided by an embodiment of the present invention;

[0047] Figure 3 A block diagram of a second sub-process of the information data collection method based on artificial intelligence provided by an embodiment of the present invention;

[0048] Figure 4 A block diagram of a third sub-process of the information data collection method based on artificial intelligence provided by an embodiment of the present invention;

[0049] Figure 5 A block diagram of the composition of an information data acquisition system based on artificial intelligence provided by an embodiment of the present invention;

[0050] Figure 6A block diagram of the components of the building module in the artificial intelligence-based information data acquisition system provided by an embodiment of the present invention;

[0051] Figure 7 A block diagram of the composition of the upload module in the artificial intelligence-based information data acquisition system provided by an embodiment of the present invention;

[0052] Figure 8 This is a block diagram of the components of the building blocks in the artificial intelligence-based information data acquisition system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0054] In Example 1, Figure 1 The implementation process of the information data collection method based on artificial intelligence provided by an embodiment of the present invention is shown and described in detail below:

[0055] S100: Build an information data collection platform, determine the registration device, select the collection unit and processing unit, identify the network service device in each processing unit, and build an edge network.

[0056] Build an information data collection platform, which is mainly used to manage registered devices and publish collection results; when the user completes registration on the information data collection platform, identify the registered device in the registration information, and the registered device refers to a device that can participate in the collection and processing of pedestrian flow information data, such as roadside cameras, monitoring equipment carried by drones, portable sensing terminals and data processing servers, etc.; from the registered devices, select the collection unit and the processing unit, wherein the collection unit is mainly responsible for the acquisition of front-end data, and the processing unit has certain computing, storage and processing capabilities. The processing unit is mainly used to perform localized analysis, feature extraction and other processing on the data uploaded by the collection unit; in this application, the collection unit is a household vehicle, and the processing unit is a roadside merchant.

[0057] From the processing units, devices with certain computing power and network connection capabilities are selected as network service devices, such as wireless gateways, smart routers and local servers. One or more network service devices in each processing unit are used to form an edge network. The edge network refers to a data network that can process the video data collected by the collection unit at the edge side.

[0058] S200: Utilize the video surveillance equipment pre-integrated in the acquisition unit to collect video data and extract segments containing roadside features. When the acquisition unit scans the service signal of the edge network, it generates an access request, establishes a communication link between the acquisition unit and the network service device, receives the usage permission issued by the user, and uploads the segment to the processing unit via the communication link.

[0059] Using the video surveillance equipment pre-integrated in the acquisition unit, monitoring video data of the family vehicle along the way is collected in real time, and the video stream is preliminarily analyzed through image processing algorithms to intercept fragments containing obvious roadside features, where the video surveillance equipment is the driving recorder in the family vehicle, and the roadside features are the sidewalk, bus stop or roadside, etc.; when the vehicle stops, the surrounding service information signals are scanned to determine whether there is a processing unit. If so, a connection request is initiated to the processing unit. Once the access request is responded to by the processing unit, a stable communication link is established between the acquisition unit and the processing unit; after receiving the user's permission to use the video data, the fragment is uploaded to the processing unit through the communication link, where the permission to use can be sent to the user in the form of an authorization pop-up window, and "Yes" and "No" buttons are opened in the authorization pop-up window.

[0060] S300: Build a pedestrian recognition model and deploy it to the processing unit. Input the fragment into the pedestrian recognition model, output the number of people in each video frame, and calculate the flow of people in each fragment. Obtain the attribute data of each fragment, where the attribute data includes at least: location and time. Write the flow of people and attribute data into the fragment, generate the collection results, and upload them to the information data collection platform.

[0061] A deep learning-based pedestrian recognition model is constructed and trained using a large number of annotated street scene pedestrian images to enable the pedestrian recognition model to recognize pedestrians in different postures and occlusion states in complex environments. The trained pedestrian recognition model is deployed to the edge processing unit. When the processing unit receives the fragment, the fragment is input into the pedestrian recognition model. By analyzing the fragment frame by frame, the pedestrian targets in the fragment are extracted, the number of pedestrians in each video frame is calculated, and the number is accumulated and counted to output the pedestrian flow in each fragment. It should be noted that the pedestrian flow is not required to be an exact number, it can be an approximate number. Some suspected pedestrian features can be filtered and not included in the pedestrian flow.

[0062] In addition, not every network service device needs to deploy a pedestrian recognition model. The network service devices of neighboring businesses can be used to jointly form an edge network, and the pedestrian recognition model can be deployed in this edge network to process the video data.

[0063] From the video surveillance equipment, the attribute data of each clip is determined, including the corresponding geographic location information (such as GPS coordinates or road signs) and time information (such as specific date, time period or timestamp) at the time of shooting; the pedestrian flow and attribute data are written into the clip, where the pedestrian flow and attribute data can be written into the clip in the form of tags or watermarks, and the collection results are generated and uploaded to the information data collection platform. The information data collection platform uniformly stores, archives, retrieves and further analyzes the collection results, thereby providing a data basis for urban management, commercial site selection and pedestrian flow prediction.

[0064] In this application, when the collection unit (vehicle) parks, the video data in the vehicle's driving recorder is sent to the network service equipment of the merchant at the parking location. The pedestrian recognition model in the merchant's wireless network is used to calculate the pedestrian flow in each segment. Based on the shooting location, time and pedestrian flow of the segment, the collection results are generated and uploaded to the information data collection platform.

[0065] In Example 2, Figure 2 The implementation process of the information data collection method based on artificial intelligence provided by an embodiment of the present invention is shown. The following details the steps of determining registrants, clustering them into collection units and processing units, identifying network service devices in each processing unit, and building an edge network.

[0066] S101: Configuring evaluation indicators of a processing unit, including at least equipment utilization and transmission delay.

[0067] Determine the evaluation indicators of the network service equipment in each processing unit, where the evaluation indicators include: equipment utilization and transmission delay. The evaluation indicators are mainly used to evaluate the excellence of each processing unit.

[0068] S102: Based on the evaluation index, a guidance mechanism is created with the collection unit as the object.

[0069] The guidance mechanism is to guide the collection unit to the processing unit with better performance; for example, according to the evaluation indicators of network service equipment, network service equipment is divided into three excellence levels: high, medium and low. The higher the excellence level, the stronger the comprehensive service capability of the network service equipment, including higher computing performance, larger storage capacity, more stable network connection and lower latency response capability, etc.; the excellence level of each network service equipment can be published to the information data collection platform for users to select the best processing unit for parking and optimize resource scheduling.

[0070] In Example 3, Figure 2 The implementation process of the information data collection method based on artificial intelligence provided by an embodiment of the present invention is shown. The steps of building the edge network are described in detail as follows:

[0071] S103: Calculate the number of concurrent connections of each edge network, select a set value, and generate an avoidance reminder based on the attribute data when the number of concurrent connections is greater than the set value.

[0072] Calculate the number of concurrent connections for each edge network, where the number of concurrent connections is the number of collection units that the network service device in the edge network maintains connection. Select a set value. When the number of concurrent connections is greater than the set value, generate an avoidance reminder. For example, the avoidance reminder can be: "The edge network load at the current parking point is large. It is recommended to choose a nearby parking point."

[0073] S104: Building a coverage range according to the service signal, and sending the avoidance reminder to the collection unit when the collection unit enters the coverage range.

[0074] The coverage range of each edge network is determined based on the reception range of the service signal. When the collection unit enters the coverage range, the avoidance reminder is sent to the corresponding driver terminal through the registration information of the collection unit in the information data collection platform, thereby reminding the driver to adjust the parking point in time.

[0075] In Example 4, Figure 3 The implementation process of the information data collection method based on artificial intelligence provided by an embodiment of the present invention is shown. The following details the steps of using the video surveillance equipment pre-integrated in the collection unit to collect video data and extract segments containing roadside features, as follows:

[0076] S201: Delineate the collection range of information data and draw a planar road network map within the collection range.

[0077] Determine the collection scope of information data, where the collection scope is determined by the administrator of the information data collection platform. The collection scope can be a certain administrative district or several streets. Obtain road network information within the collection scope from map service providers or public data, and draw a planar road network map.

[0078] S202: Mark the traffic light intersections, and cut out a number of road sections from the planar road network map, create a time window corresponding to each road section, and shift the time window after receiving the collection results.

[0079] Mark traffic light intersections on the planar road network map, divide each road in the planar road network map into multiple sections, and set a time window for each section. When the collection results are received, the time window is offset.

[0080] For example, there are three sections A, B and C in the plane road network (actually there should be more). Collection unit A travels through section A at 10:10. Assuming that the time windows of the three sections are unified as 9:00-10:00, at this time, by analyzing the video data from the driving recorder in vehicle A, the pedestrian flow data of section A is updated, and the time window corresponding to section A is offset to: 10:00-11:00. However, since there is no updated data for sections B and C, the time window remains 9:00-10:00.

[0081] In Example 5, Figure 4 The implementation process of the information data collection method based on artificial intelligence provided by an embodiment of the present invention is shown. The steps of constructing a pedestrian recognition model, embedding it into a processing unit, and inputting the fragment into the pedestrian recognition model are described in detail below:

[0082] S301: Inserting a target detection module into the pedestrian recognition model, collecting historical video data, and annotating it to generate a training set, and training the target detection module.

[0083] S302: Input the segments into a target detection module in parallel, and output a multi-target feature, wherein the multi-target feature at least includes: roadside vendors and roadside vehicles.

[0084] A target detection module is inserted into the original pedestrian recognition model as a pre-component. The target detection module is mainly used to accurately identify multi-target features in the video image, including roadside vendors, illegally parked vehicles, etc.; historical video data is collected, and the multi-target features therein are marked to generate a training set, which is used to train the target detection module.

[0085] In this application, in addition to using the pedestrian recognition model to identify pedestrians from video data, the target detection module can also be used to identify roadside vendors and roadside vehicles parked in non-designated areas in the video footage from the video data. The target detection module is built based on deep learning technology and can extract and compare features such as personnel posture, stall structure and vehicle characteristics, thereby determining road-occupying businesses and illegally parked vehicles, etc., providing a reference for law enforcement by urban management, traffic management and other departments.

[0086] In Example 6, different from Example 1, in this embodiment of the present invention, the method further includes:

[0087] The segments where the multi-target features are located are defined as target segments, the target segments and attribute data are integrated to generate a roadside information set, and then uploaded to the preset platform;

[0088] Through the information data collection platform, the user's positioning permission is obtained, the location of the target segment is used as the center of the circle, the preset distance is used as the radius, the verification range is defined, the verification task is generated, and it is published to the information data collection platform.

[0089] A fragment containing multiple target features is defined as a target segment. In other words, the target segment is the video data containing roadside vendors occupying the road and illegally parked vehicles; a correspondence between the target segment and the attribute data is established, all the target segments are integrated, and a roadside information set is generated. The roadside information set is uploaded to a preset platform, where the preset platform is a public participation platform for urban governance, for example, Guangzhou's "Snap Shot" and Shenzhen's "City Eye" in the existing technology.

[0090] After the target segment is uploaded, the verification range is defined with the target segment's location as the center and the preset distance as the radius, and a verification task is generated. For example, if the target segment is Section B, after the target segment is uploaded to the preset platform for the first time and after a period of time (such as a few hours or 1 day), a verification task is issued to users within the verification range of Section B, guiding users to Section B to capture roadside video data and determine whether situations such as roadside vendors occupying the road and illegally parked vehicles have been eliminated.

[0091] In Example 7, different from Example 1, in this embodiment of the present invention, the method further includes:

[0092] Establish the corresponding relationship between the flow of people and the time window, and select the middle point from the time window;

[0093] Use the middle point as the horizontal coordinate and the corresponding flow of people as the vertical coordinate to draw a trend graph.

[0094] Count the pedestrian flow in each time window and establish a corresponding relationship between the pedestrian flow and the time window; determine the middle moment of the time window and use it as the horizontal coordinate, and use the pedestrian flow in the corresponding time window as the vertical coordinate to draw a change trend graph; this change trend graph can intuitively show the density of pedestrian flow in the area, which is helpful for analyzing peak periods, trough periods and the periodic laws of pedestrian flow changes.

[0095] It should be noted that due to the randomness of the time window update, the update frequency of the pedestrian flow data for each road section is not the same; for example, section C may have multiple collection units (greater than the threshold) passing by every hour, and the update frequency is also multiple times per hour, but section D may have only a small number of collection units (less than or equal to the threshold) passing by every day. In this case, a change trend chart for section D is not drawn because, according to common sense, a road section with only a few vehicles passing by every day must have a small pedestrian flow, and the corresponding road section has no analysis value.

[0096] Figure 5The following is a structural block diagram of an information data acquisition system based on artificial intelligence according to an embodiment of the present invention. The information data acquisition system based on artificial intelligence 1 includes:

[0097] Building module 11, used to build an information data collection platform, determine registrants, cluster them into collection units and processing units, identify network service devices in each processing unit, and build an edge network;

[0098] The upload module 12 is configured to utilize the video surveillance equipment pre-integrated in the acquisition unit to collect video data and extract segments containing roadside features. When the acquisition unit scans the service signal of the edge network, it generates an access request, establishes a communication link between the acquisition unit and the network service device, receives the usage permission issued by the user, and uploads the segments to the processing unit via the communication link.

[0099] The construction module 13 is used to construct a pedestrian recognition model, which is embedded in the processing unit. The fragment is input into the pedestrian recognition model, the number of people in each video frame is output, and the flow of people in each fragment is calculated by superposition, and the attribute data of each fragment is obtained, wherein the attribute data at least includes: location and time. The flow of people and attribute data are written into the fragment, and the collection results are generated and uploaded to the information data collection platform.

[0100] Figure 6 The following is a structural block diagram of the information data acquisition system based on artificial intelligence provided by an embodiment of the present invention. The building module 11 includes:

[0101] The configuration unit 111 is used to configure the evaluation index of the processing unit, including at least: device utilization and transmission delay;

[0102] A creation unit 112, configured to create a guidance mechanism with the collection unit as the object according to the evaluation index;

[0103] a calculation unit 113 configured to calculate the number of concurrent connections of each edge network, select a set value, and generate an avoidance reminder based on the attribute data when the number of concurrent connections is greater than the set value;

[0104] The sending unit 114 is configured to build a coverage area according to the service signal, and send the avoidance reminder to the collection unit when the collection unit enters the coverage area.

[0105] Figure 7 The following is a structural block diagram of the information data acquisition system based on artificial intelligence provided by an embodiment of the present invention. The upload module 12 includes:

[0106] A delineation unit 121 is used to delineate a collection range of information data and draw a planar road network map within the collection range;

[0107] The offset unit 122 is used to mark the traffic light intersection, cut out a number of road sections from the planar road network map, create a time window corresponding to each road section, and offset the time window after receiving the collection result.

[0108] Figure 8 The following is a structural block diagram of the information data acquisition system based on artificial intelligence provided by an embodiment of the present invention. The building block 13 includes:

[0109] A training unit 131 is configured to insert a target detection module into the pedestrian recognition model, collect and annotate historical video data, generate a training set, and train the target detection module;

[0110] The output unit 132 is configured to input the segments into a target detection module in parallel, and output a multi-target feature, wherein the multi-target feature includes at least roadside vendors and roadside vehicles.

[0111] The building module 11 is mainly used to complete step S100, the uploading module 12 is mainly used to complete step S200, and the construction module 13 is mainly used to complete step S300;

[0112] The configuration unit 111 is mainly used to complete step S101, the creation unit 112 is mainly used to complete step S102, the calculation unit 113 is mainly used to complete step S103, and the issuing unit 114 is mainly used to complete step S104;

[0113] The demarcation unit 121 is mainly used to complete step S201, and the offset unit 122 is mainly used to complete step S202;

[0114] The training unit 131 is mainly used to complete step S301, and the output unit 132 is mainly used to complete step S302.

[0115] The technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above-mentioned embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0116] The above-described embodiments merely illustrate several implementations of the present invention, and while their descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.

[0117] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An information data collection method based on artificial intelligence, characterized in that: The method comprises: Build an information data collection platform, identify the registered devices, select the collection unit and processing unit, identify the network service device in each processing unit, and build an edge network; the collection unit is the household vehicle, and the processing unit is the roadside merchant; The video surveillance equipment pre-integrated in the acquisition unit is used to collect video data and extract segments containing roadside features. When the acquisition unit scans the service signal of the edge network, it generates an access request and establishes a communication link between the acquisition unit and the network service device. receiving the usage permission issued by the user, and uploading the clip to the processing unit via the communication link; Construct a pedestrian recognition model and deploy it to the processing unit. Input the clip into the pedestrian recognition model, output the number of people in each video frame, and calculate the flow of people in each clip by superposition. Obtain attribute data of each clip, wherein the attribute data includes at least: location and time; Write the traffic and attribute data into the fragment, generate the collection results, and upload them to the information data collection platform; The steps of determining the registration device, selecting the collection unit and the processing unit, identifying the network service device in each processing unit, and building the edge network include: Configure the evaluation indicators of the processing unit, including at least: equipment utilization and transmission delay; Based on the evaluation indicators, a guidance mechanism is created with the collection unit as the object; The steps of building the edge network include: Calculating the number of concurrent connections of each edge network, selecting a set value, and generating an avoidance reminder based on the attribute data when the number of concurrent connections is greater than the set value; Building a coverage area according to the service signal, and sending the avoidance reminder to the collection unit when the collection unit enters the coverage area; The steps of constructing a pedestrian recognition model, deploying it to a processing unit, and inputting the fragments into the pedestrian recognition model include: Inserting a target detection module into the pedestrian recognition model, collecting historical video data, annotating it, generating a training set, and training the target detection module; Inputting the segments into a target detection module in parallel, and outputting a multi-target feature, wherein the multi-target feature at least includes: roadside vendors and roadside vehicles; The method further comprises: The segments where the multi-target features are located are defined as target segments, the target segments and attribute data are integrated to generate a roadside information set, and then uploaded to the preset platform; Through the information data collection platform, the user's positioning permission is obtained, the location of the target segment is used as the center of the circle, the preset distance is used as the radius, the verification range is defined, the verification task is generated, and it is published to the information data collection platform.

2. The information data collection method based on artificial intelligence according to claim 1 is characterized in that: The step of using the video surveillance equipment pre-integrated in the acquisition unit to collect video data and extracting a segment containing roadside features includes: Define the collection scope of information data and draw a planar road network map within the collection scope; Traffic light intersections are marked, and several road sections are cut out from the planar road network map. A time window corresponding to each road section is created. When the collection results are received, the time window is shifted.

3. The information data collection method based on artificial intelligence according to claim 2 is characterized in that: The method further comprises: Establish the corresponding relationship between the flow of people and the time window, and select the middle point from the time window; Use the middle point as the horizontal coordinate and the corresponding flow of people as the vertical coordinate to draw a trend graph.

Images