A smart city traffic signal light time setting method and Internet of Things system

Abstract

The embodiments of the present specification provide a time setting method for a smart city traffic light and an Internet of Things system. The method is implemented by the time setting Internet of Things system for a smart city traffic light. The Internet of Things system includes a user platform, a service platform, a management platform, a sensor network platform and an object platform. The method is executed by the management platform and includes: obtaining pedestrian information and intersection information of a target intersection based on the object platform through the sensor network platform, where the target intersection is an intersection with a traffic light; determining a time setting scheme for the traffic light at the target intersection based on the pedestrian information and the intersection information; and sending a control instruction corresponding to the time setting scheme to the object platform through the sensor network platform, and the object platform responds to the received control instruction to control the lighting duration of the traffic light.

Description

Technical Field

[0001] The present invention relates to the field of traffic lights, and in particular to a time setting method for smart city traffic lights and an Internet of Things system. Background Art

[0002] Traffic lights play an important role in the orderly operation of urban traffic. The normal operation of traffic lights usually affects the traffic conditions at various intersections in the city.

[0003] Therefore, it is hoped to provide a time setting method and system for traffic lights in a smart city, which can optimize the time setting of traffic lights, thereby maximizing the passing rate of vehicles and / or people, avoiding traffic congestion, and ensuring smooth urban traffic. Summary of the invention

[0004] One or more embodiments of the present specification provide a method for setting the time of a smart city traffic light, the method being implemented by a time setting Internet of Things system for a smart city traffic light, the time setting Internet of Things system for a smart city traffic light comprising a user platform, a service platform, a management platform, a sensor network platform and an object platform, the method being executed by the management platform, the method comprising: obtaining pedestrian information and intersection information of a target intersection based on an object platform through a sensor network platform, the target intersection being an intersection where a traffic light is set; determining a time setting scheme for the traffic light at the target intersection based on the pedestrian information and the intersection information; and sending a control instruction corresponding to the time setting scheme to the object platform through the sensor network platform, the object platform responding to the received control instruction to control the lighting duration of the traffic light.

[0005] One or more embodiments of the present specification provide an Internet of Things system for setting the time of traffic lights in a smart city, the Internet of Things system comprising a user platform, a service platform, a management platform, a sensor network platform and an object platform, wherein the management platform is configured to perform the following operations: obtaining pedestrian information and intersection information of a target intersection based on the object platform through the sensor network platform, the target intersection being an intersection with traffic lights; determining a time setting scheme for the traffic lights at the target intersection based on the pedestrian information and intersection information; and sending control instructions corresponding to the time setting scheme to the object platform through the sensor network platform, and the object platform responds to the received control instructions to control the lighting duration of the traffic lights.

[0006] One or more embodiments of the present specification provide a computer-readable storage medium, wherein the storage medium stores computer instructions. When a computer reads the computer instructions in the storage medium, the computer executes a method for setting the time of a smart city traffic light. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] This specification will be further described in the form of exemplary embodiments, which will be described in detail by the accompanying drawings. These embodiments are not restrictive, and in these embodiments, the same number represents the same structure, wherein:

[0008] Figure 1 is an exemplary schematic diagram of a time setting IoT system for smart city traffic lights according to some embodiments of this specification;

[0009] Figure 2 is an exemplary flow chart of a method for setting the time of a smart city traffic signal light according to some embodiments of this specification;

[0010] Figure 3 is an exemplary flow chart based on determining the current number of pedestrians according to some embodiments of this specification;

[0011] Figure 4 is an exemplary schematic diagram of determining the current number of pedestrians based on a pedestrian recognition model and a first pedestrian prediction model according to some embodiments of this specification;

[0012] Figure 5 is an exemplary schematic diagram of determining the number of pedestrians at multiple future time points by using a second pedestrian prediction model according to some embodiments of this specification;

[0013] Figure 6 It is an exemplary schematic diagram of a time setting scheme for determining a traffic light at a target intersection through vector matching according to some embodiments of this specification. DETAILED DESCRIPTION

[0014] In order to more clearly illustrate the technical solutions of the embodiments of this specification, the following is a brief introduction to the drawings required for the description of the embodiments. Obviously, the drawings described below are only some examples or embodiments of this specification. For ordinary technicians in this field, this specification can also be applied to other similar scenarios based on these drawings without creative work. Unless it is obvious from the language environment or otherwise explained, the same reference numerals in the figures represent the same structure or operation.

[0015] It should be understood that the "system", "device", "unit" and / or "module" used herein are a method for distinguishing different components, elements, parts, portions or assemblies at different levels. However, if other words can achieve the same purpose, the words can be replaced by other expressions.

[0016] As shown in this specification and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "comprise" and "include" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0017] Flowcharts are used in this specification to illustrate the operations performed by the system according to the embodiments of this specification. It should be understood that the preceding or following operations are not necessarily performed precisely in order. Instead, the steps may be processed in reverse order or simultaneously. At the same time, other operations may also be added to these processes, or one or more operations may be removed from these processes.

[0018] Figure 1 This is an exemplary schematic diagram of an Internet of Things system for setting the time of traffic lights in a smart city according to some embodiments of this specification.

[0019] It should be understood that the time setting IoT system for smart city traffic lights can be implemented in various ways. Figure 1 As shown, the time setting IoT system for smart city traffic lights includes a user platform, a service platform, a management platform, a sensor network platform and an object platform. In some embodiments, the time setting IoT system for smart city traffic lights can be a part of a processing device or implemented by a processing device.

[0020] The user platform may refer to a user-led platform. In some embodiments, the user platform may be configured as a terminal device. In some embodiments, the user platform may obtain the user's input instructions through the terminal to query the time setting scheme of the traffic lights at each intersection in each area of ​​the city. In some embodiments, the urban area may be determined according to the administrative area of ​​the city, or may be flexibly determined according to preset rules, real-time data of traffic flow, etc. Each urban area may include various intersections within the area. For example, during rush hour and rush hour, the corresponding sections of all commercial office areas in the city may be divided into the same area. For another example, according to the real-time situation of congestion, all congested sections in the city may be divided into the same area. In some embodiments, the user platform may issue a time setting scheme query instruction to the service platform, receive the time setting scheme uploaded by the service platform, and feed it back to the user.

[0021] A service platform refers to a platform that can provide input and output services to users. In some embodiments, the service platform can be arranged in an independent manner. An independent arrangement means that the platform includes multiple independent sub-platforms, and the multiple sub-platforms operate and process data independently of each other, and directly interact with the upper and lower functional platforms for data. In some embodiments, the service platform can be divided into multiple independent sub-platforms according to urban areas. In some embodiments, the sub-platform of the service platform can receive a time setting scheme query instruction issued by the user platform, and pass the time setting scheme query instruction to the management platform. The sub-platform of the service platform can also receive the time setting scheme uploaded by the management platform, and pass the time setting scheme to the user platform, etc.

[0022] The management platform may refer to a platform that coordinates the connection and collaboration between various functional platforms and gathers all the information of the IoT system for setting the time of smart city traffic lights. The management platform may provide perception management and control management functions for the operation system of the IoT system for setting the time of smart city traffic lights. In some embodiments, the management platform may be a remote platform controlled by management personnel, artificial intelligence, or preset rules. In some embodiments, the management platform may be arranged independently. The management platform may be provided with a general database and multiple sub-platforms, each of which may include its own database. In some embodiments, the management sub-platform may correspond to each urban area and correspond one-to-one with the service sub-platform. Multiple sub-platforms respectively store and process data from different urban areas sent by the sensor network platform. The general database aggregates and stores and processes the data of multiple sub-platforms. In some embodiments, the management sub-platform may receive a query instruction for the time setting scheme issued by the corresponding service sub-platform, and upload the time setting scheme to the service sub-platform. The management sub-platform may also issue an instruction to query pedestrian information and intersection information to the sensor network platform, and receive the pedestrian information and intersection information uploaded by the sensor network platform for analysis and processing to determine the time setting scheme.

[0023] The sensor network platform may refer to a functional platform for managing sensory communications. The sensor network platform may realize the functions of sensing information sensory communications and controlling information sensory communications. In some embodiments, the sensor network platform may be arranged independently. The sensor network platform may include multiple independent sub-platforms. Each sensor network sub-platform may correspond to an urban area and correspond one-to-one with a management sub-platform. In some embodiments, the sensor network platform may be configured as a communication network and a gateway device, and each sensor network sub-platform may be configured as an independent gateway device. In some embodiments, the sensor network sub-platform may obtain pedestrian information and intersection information at an intersection from the object platform of the corresponding area. Each sensor network sub-platform may upload pedestrian information and intersection information of each area to the corresponding management sub-platform.

[0024] The object platform may refer to a functional platform for generating sensory information. In some embodiments, the object platform may be configured as various types of devices to obtain information. Among them, various types of devices may include devices such as traffic lights, image acquisition devices, and pedestrian buttons. Each device is configured with a unique identifier and can be used to be deployed at intersections in different areas to obtain pedestrian information and intersection information at each intersection. In some embodiments, the object platform may include several object sub-platforms, and each object sub-platform may correspond to the aforementioned types of devices in an area. In some embodiments, the object platform sub-platform may receive instructions for obtaining pedestrian information and intersection information at each intersection in the area issued by the corresponding sensor network sub-platform, and upload pedestrian information and intersection information at each intersection in the area to the corresponding sensor network sub-platform. In some embodiments, the object platform sub-platform may control the lighting duration of the traffic light in response to the control instruction corresponding to the received time setting scheme. In some embodiments, the object platform may also directly adjust the lighting duration of the traffic light or switch the traffic light based on the pedestrian button configured thereon.

[0025] Each "service sub-platform-management sub-platform-sensor network sub-platform-object sub-platform" can be an independent branch, which processes the relevant data of each intersection in the corresponding urban area in parallel and independently (such as pedestrian information, intersection information, time setting plan, etc.), and independently feeds back to the user platform.

[0026] In some embodiments, the IoT system for setting the time of traffic lights in smart cities can be applied to various traffic scenarios such as railways, roads, waterways, and aviation to determine various work plans including the time setting of traffic lights, traffic control, and the deployment of relevant personnel. It should be noted that the above scenarios are only examples and do not limit the specific application scenarios of the IoT system for setting the time of traffic lights in smart cities. Those skilled in the art can apply the IoT system for setting the time of traffic lights in smart cities to any other appropriate scenarios based on the content disclosed in this embodiment.

[0027] The following will take the application scenario of the IoT system for setting the time of traffic lights in a smart city as an example to specifically explain the IoT system for setting the time of traffic lights in a smart city.

[0028] The management platform can be configured to obtain pedestrian information and intersection information of a target intersection based on the object platform through the sensor network platform, where the target intersection is an intersection with a traffic light; determine a time setting scheme for the traffic light at the target intersection based on the pedestrian information and intersection information; and send a control instruction corresponding to the time setting scheme to the object platform through the sensor network platform, and the object platform responds to the received control instruction to control the lighting duration of the traffic light. For more information on determining the time setting scheme, please refer to Figure 2and its related description.

[0029] In some embodiments, the management platform may also be configured to obtain a target image of a target intersection through the object platform; and determine pedestrian information at the target intersection based on the image recognition result of the target image. For more information about determining pedestrian information at a target intersection through a target image, please refer to Figure 2 and its related description.

[0030] In some embodiments, the object platform includes a pedestrian button disposed near a crosswalk, and the pedestrian button is used to adjust the lighting duration of the traffic light. For more information about the pedestrian button, please refer to Figure 2 and its related description.

[0031] In some embodiments, the target image includes at least one frame of real-time image and multiple frames of historical images. Pedestrian information includes the current number of pedestrians. The management platform is configured to determine a first number of pending pedestrians based on at least one frame of real-time image through a pedestrian recognition model, where the pedestrian recognition model is a machine learning model. The management platform determines a second number of pending pedestrians based on multiple frames of historical images through a first pedestrian prediction model, where the first pedestrian prediction model is a machine learning model. The management platform determines the current number of pedestrians based on the first number of pending pedestrians and the second number of pending pedestrians. For more information on determining the current number of pedestrians, see Figure 3 and its related description.

[0032] In some embodiments, the management platform may also be configured to determine the current number of pedestrians based on a weighted sum of the first number of pedestrians to be determined and the second number of pedestrians to be determined, wherein the weights of the first number of pedestrians to be determined and the second number of pedestrians to be determined are determined based on the clarity of the real-time image and the historical image, respectively, and the weights are positively correlated with the clarity. For more information on the weighted sum of the first number of pedestrians to be determined and the second number of pedestrians to be determined, see Figure 3 and its related description.

[0033] In some embodiments, the first pedestrian prediction model can be obtained by training, where the training samples are multiple historical images at a first sample time, the labels are the actual number of pedestrians at a second sample time, and the first sample time is earlier than the second sample time. For more information on training the first pedestrian prediction model, see Figure 3 and its related description.

[0034] In some embodiments, the pedestrian information also includes the number of pedestrians at multiple future time points. For more information about the number of pedestrians at multiple future time points, see Figure 2 and its related description.

[0035] In some embodiments, the management platform can also be configured to determine the number of pedestrians at multiple future time points through a second pedestrian prediction model based on multiple frames of historical images of the target screen, intersection information, event information, weather information, and target time period. The second pedestrian prediction model is a machine learning model. For more information about the second pedestrian prediction model, see Figure 5 and its related description.

[0036] In some embodiments, the management platform may also be configured to determine the update frequency of the time setting scheme. For more information on determining the update frequency, see Figure 2 and its related description.

[0037] In some embodiments, the management platform can also be configured to determine the time setting scheme of the traffic light at the target intersection through vector matching based on the pedestrian information and the intersection information. For more information about determining the time setting scheme through vector matching, please refer to Figure 6 and its related description.

[0038] It should be noted that the above description of the IoT for setting the time of traffic lights in smart cities is only for the convenience of description and cannot limit this specification to the scope of the embodiments. It is understandable that for those skilled in the art, after understanding the principle of the system, they may arbitrarily combine the modules or form a subsystem to connect with other modules without deviating from this principle. For example, Figure 1 The management platform, user platform, service platform, sensor network platform and object platform disclosed in the system may be different platforms in one system, or one platform may realize the functions of two or more of the above platforms.

[0039] Figure 2 FIG. 1 is an exemplary flow chart of a method for setting the time of a smart city traffic signal light according to some embodiments of this specification. Figure 2 As shown, the process 200 includes the following steps. In some embodiments, the process 200 may be executed by a management platform.

[0040] Step 210, obtaining pedestrian information and intersection information of a target intersection based on an object platform through a sensor network platform, where the target intersection is an intersection with a traffic light.

[0041] The target intersection is an intersection with a traffic light, such as a crossroads or an intersection near a school.

[0042] Pedestrian information refers to information reflecting the situation of pedestrians at the intersection, for example, the number of pedestrians (e.g., 354 people).

[0043] In some embodiments, the management platform can obtain pedestrian information at the target intersection based on the object platform through the sensor network platform. For example, the object platform collects pedestrian information based on an information collection terminal (such as an image acquisition device, etc.). The management platform can issue an instruction to obtain intersection information to the sensor network platform. The sensor network platform issues the instruction to the object platform. The object platform responds to the instruction and sends the intersection information to the sensor network platform. The management platform can receive pedestrian information uploaded by the sensor network platform.

[0044] In some embodiments, the management platform may obtain a target image of a target intersection through the object platform; and determine pedestrian information at the target intersection based on an image recognition result of the target image.

[0045] The target image refers to an image (such as an RGB image, a grayscale image, an infrared image, etc.) used to determine pedestrian information at the target intersection. The target image may contain one or more pedestrians at the target intersection.

[0046] In some embodiments, the target image includes at least one real-time image and multiple historical images. For example, if the current time is 15:00, the target image may include an image corresponding to 15:00 and images corresponding to 14:55, 14:56, 14:57, 14:58 and 14:59, respectively.

[0047] In some embodiments, the object platform may include an image acquisition device (eg, a surveillance camera, a drone, etc.) configured at an intersection. The management platform may collect target images based on the image acquisition device on the object platform through the sensor network platform.

[0048] In some embodiments, the object platform may further include a pedestrian button disposed near a crosswalk. The pedestrian button may be used to adjust the duration of a traffic light (e.g., extend the duration of a green light by 5 seconds) or switch a traffic light (e.g., switch a red light to a green light or vice versa). When a pedestrian is about to cross a crosswalk, the pedestrian may press the pedestrian button to dynamically adjust the duration of a traffic light or switch a traffic light.

[0049] In some embodiments, the pedestrian button can be used at certain specific intersections, for example, intersections with fewer pedestrians but larger traffic volume. It is understandable that if a crosswalk at a certain intersection is unmanned for a long time, there is no need to alternate the traffic lights to facilitate vehicle traffic. When a pedestrian needs to cross the crosswalk, the pedestrian can press the pedestrian button to turn the traffic light green. In some embodiments, after the pedestrian presses the pedestrian button, the traffic light can be switched after a first preset time (e.g., 2 seconds).

[0050] In some embodiments, the management platform can also determine the target screen by screening multiple frames of images in the initial screen. The initial screen refers to a continuous screen directly collected by the image acquisition device configured by the object platform. The management platform can select image frames at intervals of a preset time length in the initial screen, and use the selected image frames as target frames, and form the target screen in chronological order. Therefore, the time interval between two adjacent target frames in the initial screen is a preset time length. For example, the initial screen has a continuous screen of 30 seconds. The preset time length is 2 seconds. The management platform can use the image frames corresponding to the 0th second, the 2nd second, the 4th second... the 30th second as target frames to form the target screen. The time interval between two adjacent target frames in the initial screen is 2 seconds. In screening the initial screen, when the multiple frames of images collected are closer to the time point when the pedestrian button is pressed, the management platform can shorten the time interval between the two adjacent target frames selected in the initial screen. In other words, the closer the images collected are to the time point when the pedestrian button is pressed, the more target frames the management platform selects.

[0051] As an example only, the management platform can set a preset time interval between two adjacent target frames in the initial screen. Within the second preset time range (e.g., 0 to 1 second) before and after the time when the pedestrian button is pressed, the time interval between two adjacent target frames in the initial screen can be reduced by 0.02 seconds based on the preset time interval. Within the third preset time range (e.g., 1 to 2 seconds) before and after the time when the pedestrian button is pressed, the preset time interval between two adjacent target frames in the initial screen can be reduced by 0.01 seconds based on the preset time interval. For example, the preset time interval between two adjacent target frames in the initial screen can be 0.03 seconds. The time when the pedestrian presses the button is 9:00:00. Then within the second preset time range, that is, within 8:59:59 to 9:00:01, the time interval between two adjacent target frame images in the initial screen is 0.01 seconds. In the third preset time range, that is, within 8:59:58-8:59:59 and 9:00:01-9:00:02, the time interval between two adjacent target frames in the initial picture is 0.02 seconds.

[0052] The method described in some embodiments of the present specification adopts a setting that the time interval between two adjacent target frames in the initial picture is shortened to obtain the target picture when the time point when the pedestrian button is pressed is closer. Compared with taking all image frames as the target picture, this setting can collect the situation when there are pedestrians at the intersection in a targeted manner. While ensuring the comprehensiveness of effective information, it reduces the amount of data analysis so that pedestrian information can be determined more quickly and accurately.

[0053] In some embodiments, the method described in some embodiments can more flexibly switch traffic lights or adjust the lighting duration by setting a pedestrian button, so as to realize the normal passage of pedestrians or vehicles in both cases of people or no people passing through the crosswalk, and avoid unnecessary waiting of pedestrians and vehicles.

[0054] In some embodiments, the management platform can determine the pedestrian information at the target intersection based on the image recognition result of the target image. The image recognition result can be obtained by analyzing and processing the target image through algorithms such as convolutional neural network (CNN), Fisherfaces, HaarCascade, etc.

[0055] The methods described in some embodiments of this specification perform image recognition on the target screen to determine more accurate pedestrian information, thereby facilitating the subsequent determination of a more reasonable time setting plan and improving the passing rate of pedestrians and vehicles at the intersection.

[0056] In some embodiments, the pedestrian information may include the current number of pedestrians, which refers to the number of pedestrians at the target intersection at the current time.

[0057] In some embodiments, the management platform can perform face recognition based on the collected target image to obtain the current number of pedestrians.

[0058] In some embodiments, the management platform can determine a first number of pedestrians to be determined based on at least one frame of real-time image by using a pedestrian recognition model; determine a second number of pedestrians to be determined based on multiple frames of historical image by using a first pedestrian prediction model, and determine the current number of pedestrians based on the first number of pedestrians to be determined and the second number of pedestrians to be determined. For more information about determining the current number of pedestrians based on the first number of pedestrians to be determined and the second number of pedestrians to be determined, please refer to Figure 3 and its related description.

[0059] In some embodiments, the pedestrian information may include the number of pedestrians at multiple time points in the future. For example, the number of pedestrians at 18:00 in the next two days will be 230 and 245. The number of pedestrians at 14:00 and 14:30 in the afternoon of the same day will be 23 and 56.

[0060] In some embodiments, the management platform may determine the number of pedestrians at multiple future time points using a variety of methods such as preset rules, various data analysis algorithms, and / or manual experience. For example, the management platform may generate corresponding rules for historical time points and the number of pedestrians at historical time points based on the historical time points of the target intersection and the historical number of pedestrians at the historical time points in advance. Accordingly, the management platform may determine the historical number of pedestrians corresponding to the historical time points that are the same as or close to the future time points as the number of pedestrians at the current target intersection at multiple future time points based on the future time points and the corresponding rules.

[0061] In some embodiments, the management platform can determine the number of pedestrians at multiple future time points based on multiple frames of historical images of the target screen, intersection information, event information, and weather information through a second pedestrian prediction model, where the second pedestrian prediction model is a machine learning model. For more information about determining the number of pedestrians at multiple future time points through the second pedestrian prediction model, please refer to Figure 5 and its related description.

[0062] The methods described in some embodiments of this specification can reflect the pedestrian situation at the intersection at multiple future time points through the number of pedestrians at multiple future time points, which is convenient for determining a more appropriate time setting plan for traffic lights later.

[0063] The intersection information refers to information reflecting the relevant situation of the target intersection. In some embodiments, the intersection information may include at least one or more of the intersection type, road width, nearby building information, etc. Among them, the intersection type may reflect the shape of the intersection. For example, the intersection type may be a crossroads, a roundabout, a Y-shaped intersection, etc. The road width may be the width of at least one road where the intersection is located. For example, 20 meters, 10 meters, and 25 meters, etc. The building information near the intersection may reflect the building type and scale near the intersection, for example, a large shopping mall with a capacity of 5,000 people. An office building with a capacity of 1,000 people. A subway station with an average passenger flow of 300 people / minute, etc.

[0064] In some embodiments, the management platform may obtain intersection information based on the object platform through the sensor network platform. For example, the object platform may pre-store intersection information of each intersection. The management platform may issue an instruction to obtain intersection information to the sensor network platform. The sensor network platform issues the instruction to the object platform. The object platform responds to the instruction and sends the intersection information to the sensor network platform. The management platform may receive the intersection information uploaded by the sensor network platform.

[0065] Step 220, based on the pedestrian information and the intersection information, determine the time setting plan of the traffic light at the target intersection.

[0066] The time setting scheme refers to the lighting duration of various indicator lights in a traffic light in a lighting cycle. Among them, a lighting cycle can refer to the total duration of all signal lights being lit once, such as 30 seconds, 1 minute, etc. For example, the time setting scheme can be that the lighting duration of the red light is 40 seconds, and the lighting duration of the green light is 30 seconds. For another example, the time setting scheme can also be that the lighting duration of the red light is 30 seconds, the lighting duration of the yellow light is 5 seconds, and the lighting duration of the green light is 20 seconds.

[0067] In some embodiments, the management platform can generate a mapping relationship between historical intersection information, historical pedestrian information, and historical time setting schemes based on historical intersection information of each of the multiple intersections, multiple historical pedestrian information of the multiple intersections, and the historical time setting schemes of the corresponding traffic lights. It is understandable that the management platform can determine the historical time setting scheme corresponding to the historical intersection information and historical pedestrian information that are the same or similar to the current intersection information and pedestrian information as the current predicted passenger flow data based on the current intersection information, the current pedestrian information, and the mapping relationship.

[0068] In some embodiments, the management platform can determine the time setting scheme of the traffic light at the target intersection through vector matching based on the pedestrian information and the intersection information. For more information about determining the time setting scheme through vector matching, please refer to Figure 6 and its related description.

[0069] In some embodiments, the management platform may also determine the update frequency of the time setting scheme.

[0070] Update frequency refers to the number of updates completed within a unit of time in the time setting scheme, for example, once per minute, once per day, etc.

[0071] In some embodiments, the management platform may determine the updating frequency of the time setting plan based on the pedestrian passing rate of the time setting plan.

[0072] The pedestrian passing rate refers to the proportion of pedestrians passing through the crosswalk to the total number of pedestrians at the target intersection in a lighting cycle before the red light comes on. For example, if the total number of pedestrians is 120 and the number of pedestrians passing through the crosswalk is 45, then the pedestrian passing rate is 45 / 120=37.5%.

[0073] In some embodiments, the management platform may use the actual pedestrian passing rate of a certain time setting scheme as the pedestrian passing rate corresponding to the time setting scheme. The management platform may calculate the average of multiple pedestrian passing rates corresponding to the same time setting scheme in multiple historical lighting cycles, and use the average of the multiple pedestrian passing rates as the actual pedestrian passing rate.

[0074] In some embodiments, when the pedestrian passage rate is within a preset range, the management platform can determine the update frequency of the time setting scheme corresponding to the preset range according to the preset range of the pedestrian passage rate. Just as an example, when the pedestrian passage rate is between 0% and 30%, the update frequency of the time setting scheme can be once every 12 hours; when the pedestrian passage rate is between 30% and 60%, the update frequency of the time setting scheme can be once a day; when the pedestrian passage rate is greater than 60%, the update frequency of the time setting scheme can be once a week. It should be understood that the higher the pedestrian passage rate, the more reasonable the setting of the time setting scheme, and the lower the update frequency of the time setting scheme can be.

[0075] In some embodiments, the management platform may determine the update frequency of the time setting scheme based on the degree of change in the number of pedestrians. The degree of change in the number of pedestrians may reflect the fluctuation of the number of pedestrians. In some embodiments, the degree of change in the number of pedestrians may be represented by the standard deviation of the number of pedestrians. For example, at T 1 , T 2 ,……T n At this moment, the number of pedestrians is S 1 , S 2 , ……S n , then the degree of change in the number of pedestrians can be S 1 , S 2 , ……S n The standard deviation of .

[0076] In some embodiments, the management platform may pre-set a corresponding relationship between the degree of change in the number of pedestrians and the update frequency (e.g., a one-to-one correspondence, a quantitative relationship, etc.). The management platform may determine the update frequency of the time setting scheme based on the current degree of change in the number of pedestrians and the corresponding relationship between the degree of change in the number of pedestrians and the update frequency. It should be understood that the smaller the degree of change in the number of pedestrians, the more stable the number of pedestrians, and the lower the update frequency of the time setting scheme.

[0077] The method described in some embodiments of this specification sets the update frequency of the time setting scheme according to the actual number of pedestrians. Compared with a single change frequency, this method can quickly respond to the actual number of pedestrians at the target intersection, thereby updating the time setting scheme and avoiding congestion at the target intersection.

[0078] Step 230, the control instruction corresponding to the time setting scheme is sent to the object platform through the sensor network platform, and the object platform responds to the received control instruction to control the lighting duration of the traffic light. The control instruction may include the control information of the lighting duration in the time setting scheme.

[0079] In some embodiments, the object platform may adjust the lighting duration of the traffic light according to the control information in the control instruction in response to the received control instruction.

[0080] In some embodiments, the management platform can also adjust the lighting duration of the traffic light or switch the traffic light through the pedestrian button configured on the object platform. For more instructions on adjusting the lighting duration of the traffic light or switching the traffic light through the pedestrian button, please refer to the aforementioned step 220 and its related description.

[0081] The methods described in some embodiments of this specification obtain pedestrian information and intersection information at the target intersection to determine a suitable time setting plan to optimize the lighting duration of traffic lights, thereby improving the passing efficiency of pedestrians and vehicles at the intersection, avoiding congestion at the intersection, and ensuring smooth traffic at the intersection.

[0082] Figure 3 is an exemplary flow chart based on determining the current number of pedestrians according to some embodiments of this specification. Figure 3 As shown, the process 300 includes the following steps. In some embodiments, the process 300 may be executed by a management platform.

[0083] Step 310, based on at least one frame of real-time image, determine the first number of pedestrians to be determined through a pedestrian recognition model, where the pedestrian recognition model is a machine learning model.

[0084] The first pending number of pedestrians refers to the current number of pedestrians determined after the real-time image is recognized and processed.

[0085] The pedestrian recognition model can be used to process at least one frame of real-time image to determine the first number of pedestrians to be determined. The pedestrian recognition model can be a YOLO (You Only Look Once) model.

[0086] Figure 4 This is an exemplary schematic diagram of determining the current number of pedestrians based on the pedestrian recognition model and the first pedestrian prediction model. Figure 4 As shown, the input of the pedestrian recognition model 420 may include at least one frame of real-time image 410. The output of the pedestrian recognition model 420 may be the first number 430 of pedestrians to be determined.

[0087] In some embodiments, the parameters of the pedestrian recognition model can be obtained through training. The management platform can train the initial pedestrian recognition model based on multiple groups of first training samples with labels. The first training sample can be a sample image of a sample intersection, and the label of the first training sample can be the actual number of pedestrians in the sample image. It should be understood that the sample intersection in the first training sample should be the target intersection or other intersections with the same intersection information as the target intersection. The label can be obtained by manually marking the number of pedestrians in the sample image.

[0088] Input a plurality of first training samples into an initial pedestrian recognition model, construct a loss function based on the output of the initial pedestrian recognition model and the labels of the first training samples, and iteratively update the parameters of the initial pedestrian recognition model based on the loss function, and when the trained model meets the first preset condition, the training ends and the trained pedestrian recognition model is obtained. The first preset condition may include but is not limited to the convergence of the loss function, the loss function value is less than a preset value, or the number of training iterations reaches a threshold, etc.

[0089] Step 320, based on multiple frames of historical images, determine the second number of pedestrians to be determined through a first pedestrian prediction model, where the first pedestrian prediction model is a machine learning model.

[0090] The second pending number of pedestrians refers to the current number of pedestrians predicted after prediction processing of historical images.

[0091] The first pedestrian prediction model can be used to process multiple frames of historical images to determine the second number of pedestrians to be determined. The first pedestrian prediction model can be a convolutional neural network (CNN).

[0092] like Figure 4 As shown, the input of the first pedestrian prediction model 450 may include multiple frames of historical images 440. The output of the first pedestrian prediction model 450 may be the second to-be-determined number of pedestrians 460.

[0093] In some embodiments, the parameters of the first pedestrian prediction model can be obtained through training. The management platform can train the initial first pedestrian prediction model based on multiple sets of second training samples with labels. The second training samples can be multiple sample images of the sample intersection at multiple first sample times, and the labels of the second training samples can be the actual number of pedestrians at the sample intersection at the second sample time.

[0094] It should be understood that the first sample time should be earlier than the second sample time. The label can be obtained by manually annotating and counting the number of pedestrians in the image corresponding to the second sample time at the sample intersection. For more information about the training process, please refer to the above description of the training of the pedestrian recognition model.

[0095] Step 330 : Determine the current number of pedestrians based on the first number of pedestrians to be determined and the second number of pedestrians to be determined.

[0096] like Figure 4 As shown, the management platform may process the first pending pedestrian number 430 and the second pending pedestrian number 460 to determine the current pedestrian number 470. For example, the management platform may average the first pending pedestrian number and the second pending pedestrian number to determine the current pedestrian number.

[0097] In some embodiments, the management platform can determine the current number of pedestrians based on the weighted sum of the first number of pedestrians to be determined and the second number of pedestrians to be determined, wherein the weights of the first number of pedestrians to be determined and the second number of pedestrians to be determined are determined based on the clarity of the real-time image and the historical image, respectively, and the weights are positively correlated with the clarity. The clarity of the real-time image and the historical image can be obtained by the management platform through clarity quantitative analysis (such as Brenner function analysis method, Fourier transform analysis method, etc.). The current number of pedestrians can be calculated by formula (1):

[0098] N= p * M 1 + q *M 2 (1)

[0099] Among them, N is the current number of pedestrians, M 1 is the first number of pedestrians to be determined, M 2 is the second number of pedestrians to be determined. p and q are the weights corresponding to the first number of pedestrians to be determined and the second number of pedestrians to be determined.

[0100] In some embodiments, the management platform may pre-set that the weights corresponding to the first number of pedestrians to be determined and the second number of pedestrians to be determined satisfy a quantitative relationship with the clarity. Exemplarily, the quantitative relationship may be determined by the following formula:

[0101]

[0102] Among them, D 1 is the clarity of the real-time image, D 2 is the clarity of the historical image, k 1 and k 2 is a preset constant. 1 and k 2 It can be a system default value, an experience value, a manually preset value, etc. or any combination thereof, and can be set according to actual needs.

[0103] For example, D 1 =30, D 2 =40. The management platform can pre-set k 1 = 0.012 and k 2=0.016, according to formula (2), p=0.012*30=0.36, q=0.016*40=0.64. Therefore, the weights corresponding to the first number of pedestrians to be determined and the second number of pedestrians to be determined are 0.36 and 0.64 respectively.

[0104] In some embodiments, the management platform may also preset the weights of the first number of pedestrians to be determined and the second number of pedestrians to be determined, and form a corresponding relationship with the weights of the first number of pedestrians to be determined and the second number of pedestrians to be determined based on the difference in clarity between the real-time image and the historical image and the range of the difference. For example, when the difference in clarity between the real-time image and the historical image is between 0 and 50, and the clarity of the real-time image is greater than that of the historical image, the weights of the first number of pedestrians to be determined and the second number of pedestrians to be determined are 0.6 and 0.4, respectively. On the contrary, when the clarity of the real-time image is less than that of the historical image, the weights are 0.4 and 0.6. When the difference in clarity between the real-time image and the historical image is greater than 50, and the clarity of the real-time image is greater than that of the historical image, the weights of the first number of pedestrians to be determined and the second number of pedestrians to be determined are 0.8 and 0.2, respectively. On the contrary, when the clarity of the real-time image is less than that of the historical image, the weights are 0.2 and 0.8.

[0105] In some embodiments, the weight of the second number of pedestrians to be determined can be determined based on the pixel difference between each frame in the multi-frame historical image, and the pixel difference is negatively correlated with the weight of the second number of pedestrians to be determined. It should be understood that the greater the pixel difference between each frame in the multi-frame historical image, the greater the degree of change in the multi-frame historical image, and the lower the accuracy of the predicted second number of pedestrians to be determined. Therefore, the management platform needs to reduce the impact of the second number of pedestrians to be determined on the final result, that is, reduce the weight corresponding to the second number of pedestrians to be determined.

[0106] In some embodiments, within the fourth preset time after the button is pressed at the current time, the management platform can determine that the weight of the first pending number of pedestrians can be greater than the weight of the second pending number of pedestrians. For example, the fourth preset time is 5 minutes. The management platform can preset that within 5 minutes after the button is pressed at the current time, the weight of the first pending number of pedestrians is 0.7 and the weight of the second pending number of pedestrians is 0.3. It should be understood that when a pedestrian presses the pedestrian button, it means that there are pedestrians at the intersection. Therefore, real-time images play a more critical role in obtaining the accurate current number of pedestrians. The management platform can correspondingly increase the weight of the first pending number of pedestrians to obtain a more accurate current number of pedestrians.

[0107] Some embodiments of the present specification can determine different weights according to the difference in clarity between real-time images and historical images, thereby reducing errors and obtaining target fault information more accurately.

[0108] The methods described in some embodiments of this specification can determine a more accurate current number of pedestrians through comprehensive analysis of real-time images and multiple frames of historical images to reflect the pedestrian information at the intersection at the current time, thereby facilitating the subsequent determination of a more appropriate time setting plan for traffic lights.

[0109] Figure 5 This is an exemplary schematic diagram of determining the number of pedestrians at multiple future time points through a second pedestrian prediction model according to some embodiments of this specification.

[0110] In some embodiments, the management platform can determine the number of pedestrians at multiple time points in the future through a second pedestrian prediction model based on multiple frames of historical images of the target screen, intersection information, event information, weather information and target time period, and the second pedestrian prediction model is a machine learning model.

[0111] The target time period refers to the time period for which the number of pedestrians in the future needs to be determined. The target time period can be at least a period of time in the future. For example, the next 2 hours, the rush hour of each day in the next week, 18:00-19:00 on June 30, 2023, etc. In some embodiments, the target time period can include multiple time points in the future.

[0112] Event information refers to information related to large-scale events (such as exhibitions, concerts, etc.) near the intersection during the target time period. Event information may include information such as the time of the event (such as 15:00 on September 21), the location of the event (such as a stadium), and the scale of the event (such as 3,000 people). In some embodiments, event information can be determined through public event information.

[0113] Weather information refers to information about weather conditions near the intersection during a target time period, such as heavy rain, light rain, snow, sunny days, etc. In some embodiments, the weather information may be determined based on a weather forecast.

[0114] The second pedestrian prediction model can be used to analyze multiple frames of historical images, intersection information, event information, weather information and target time period to determine the number of pedestrians at multiple time points in the future. The second pedestrian prediction model can be a combination of one or more of a Transformer model, a convolutional neural network model, a deep neural network model, etc.

[0115] like Figure 5 As shown, the input of the second pedestrian prediction model 560 may include multiple frames of historical images 510, intersection information 520, event information 530, weather information 540 and target time period 550, and the output may be the number of pedestrians 570 at multiple time points in the future.

[0116] In some embodiments, the parameters of the second pedestrian prediction model can be obtained through training. The management platform can train the initial second pedestrian prediction model based on multiple groups of third training samples with labels. The third training samples can be intersection information of the sample intersection, multiple sample images of the first sample time, event information of the second sample time, weather information of the second sample time, and the second sample time. The label of the third training sample can be the actual number of pedestrians at the sample intersection at the second sample time. The label can be obtained by manually annotating and counting the number of pedestrians in the image corresponding to the sample intersection at the second sample time. For more information about the training process, please refer to the aforementioned description of the training of the pedestrian recognition model.

[0117] like Figure 5 As shown, the second pedestrian prediction model 560 includes an intersection information embedding layer 560-1, an event information embedding layer 560-2 and a prediction layer 560-3.

[0118] The intersection information embedding layer can process the intersection information to determine the intersection features. The intersection features refer to the features obtained after feature extraction of the intersection information. In some embodiments, the intersection information embedding layer can be a convolutional neural network. Figure 5 As shown, the input of the intersection information embedding layer 560 - 1 may be the intersection information 520 , and the output of the intersection information embedding layer 560 - 1 may be the intersection feature 561 .

[0119] The event information embedding layer can process the event information and determine the event features. Event features refer to features obtained after feature extraction of event information. In some embodiments, the event information embedding layer can be a convolutional neural network. Figure 5 As shown, the input of the event information embedding layer 560 - 2 may be the event information 530 , and the output of the event information embedding layer 560 - 2 may be the event feature 562 .

[0120] The prediction layer can analyze and process multiple frames of historical images, intersection features, event features, weather information, and target time periods to determine the number of pedestrians at multiple time points in the future. Figure 5 As shown, the input of the prediction layer 560-3 may be the intersection feature 561 obtained based on the intersection information embedding layer 560-1, the event feature 562 obtained based on the event information embedding layer 560-2, the multiple frames of historical images 510, the weather information 540, and the target time period 550, and the output may be the number of pedestrians 570 at multiple time points in the future. In some embodiments, the prediction layer may be a combination of one or more of a convolutional neural network, a deep neural network, and the like.

[0121] In some embodiments, the intersection information embedding layer, the event information embedding layer, and the prediction layer can be obtained by jointly training the initial pedestrian recognition model based on the third training sample and the label of the third training sample.

[0122] In some embodiments, the management platform may also use the fifth preset time before the time point when the pedestrian presses the pedestrian button as the first sample time, and use the image of the fifth preset time before the time point when the pedestrian presses the pedestrian button as the sample image. The management platform may use the time point when the pedestrian presses the pedestrian button and multiple time points within the sixth preset time thereafter as the second sample time, and use the number of pedestrians corresponding to the time point when the pedestrian presses the pedestrian button and multiple time points within the sixth preset time thereafter as the label corresponding to the third training sample. It should be understood that when the pedestrian presses the pedestrian button, it indicates that there are pedestrians at the intersection. The management platform obtains training samples and labels based on the images before and after the time point when the pedestrian presses the pedestrian button, thereby improving the efficiency of obtaining training samples and labels.

[0123] In some embodiments, the label corresponding to the third training sample can be obtained by manually annotating and counting the number of pedestrians in the image corresponding to the second sample time at the sample intersection. In some embodiments, the label corresponding to the third training sample can be determined based on at least one frame of image. Among them, at least one frame of image can be determined by intermittent collection by the object platform within the second sample time, and can also be determined by selecting at least one frame from the image corresponding to the second sample time. For example, one or more frames of image corresponding to the time point when the pedestrian presses the pedestrian button. For another example, one or more frames of image corresponding to when the traffic light is green, etc. In some embodiments of the present specification, there is no need to count each frame of image, thereby saving label acquisition time and reducing label acquisition costs.

[0124] In the joint training, the management platform can input the intersection information of the sample intersection in the third training sample into the initial intersection information embedding layer, and input the event information of the second sample time in the third training sample into the initial event information embedding layer. Then, the output of the initial intersection information embedding layer, the output of the initial event information embedding layer, multiple sample images, weather information of the second sample time, and the second sample time are input into the initial prediction layer together, and a loss function is constructed based on the output and label of the initial prediction layer. The parameters of each layer in the initial second pedestrian prediction model are iteratively updated based on the loss function so that the loss function of the model meets the preset conditions, and a trained second pedestrian prediction model is obtained. The preset conditions may include but are not limited to the convergence of the loss function, the loss function value is less than the preset value, or the number of training iterations reaches a threshold.

[0125] The methods described in some embodiments of this specification use a model to analyze and process historical images and possible situations that may affect the number of pedestrians in the future (such as intersection information, event information, weather information, and target time period, etc.), and can quickly and accurately predict the number of pedestrians at multiple future time points in the target time period, thereby facilitating the subsequent determination of a time setting plan.

[0126] Figure 6 It is an exemplary schematic diagram of a time setting scheme for determining a traffic light at a target intersection through vector matching according to some embodiments of this specification.

[0127] In some embodiments, the management platform can determine the time setting scheme of the traffic light at the target intersection through vector matching based on pedestrian information and intersection information. Figure 6 As shown, an exemplary process of determining a time setting scheme based on vector matching is as follows:

[0128] In some embodiments, the management platform can construct a feature vector 620 based on the data to be tested 610. The data to be tested 610 includes pedestrian information 610-1 and intersection information 610-2. For example only, the feature vector can be represented as [a, b, c, ...], where element a can represent pedestrian information (e.g., a corresponding to 1 when the number of pedestrians is between 0 and 100, a corresponding to 2 when the number of pedestrians is between 100 and 200, ...); element b can represent the type of intersection (e.g., b corresponding to 1 when the intersection type is a crossroad, b corresponding to 2 when the intersection type is a roundabout, b corresponding to 3 when the intersection type is a Y-shaped intersection, etc.); c can represent the road width (e.g., c corresponding to 1 when the road width is 0 to 5 meters, c corresponding to 2 when the road width is 5 to 10 meters, c corresponding to 3 when the road width is 10 to 15 meters, etc.).

[0129] In some embodiments, the management platform can construct a reference vector 641 based on the historical data 630 of each intersection, and form a vector database 640 based on the reference vector 641. The method of constructing the reference vector is similar to the method of constructing the feature vector. In some embodiments, the historical data 630 can include historical pedestrian information 630-1 and historical intersection information 630-2.

[0130] In some embodiments, the management platform can also construct a feature vector 620 based on pedestrian information 610-1, intersection information 610-2, weather information 610-3, and event information 610-4. Accordingly, the management platform can construct a reference vector 641 based on historical pedestrian information 630-1, historical intersection information 630-2, historical weather information 630-3, and historical event information 630-4.

[0131] In some embodiments, each reference vector may correspond to a historical time setting scheme. It should be understood that the historical time setting scheme may be the actual lighting duration of various indicator lights of the traffic light corresponding to the historical pedestrian information, historical intersection information, etc. represented by the reference vector.

[0132] In some embodiments, the management platform may calculate a vector distance 650 between the feature vector 620 and a reference vector 641 in a vector database 640. The vector distance may be Euclidean distance, cosine distance, Mahalanobis distance, Chebyshev distance, Manhattan distance, or the like.

[0133] In some embodiments, the management platform may screen the reference vectors 641 based on the vector distance 650 to determine at least one target reference vector 660. For example, the management platform may select a reference vector with the smallest vector distance to the feature vector as the target reference vector. For another example, the management platform may use one or more reference vectors with a vector distance less than a threshold value (e.g., 2) as the target reference vector.

[0134] In some embodiments, the management platform may determine a timing scheme 670 based on the target reference vector 660 .

[0135] In some embodiments, when a target reference vector is determined, the management platform may directly use the time setting scheme corresponding to the target reference vector as the final time setting scheme.

[0136] In some embodiments, when multiple target reference vectors are determined, the management platform can average the lighting duration of the same indicator lights included in the multiple target reference vectors in the time setting schemes corresponding to the multiple target reference vectors, and use the average value as the final time setting scheme. As an example only, the management platform determines the target reference vector B and the target reference vector C. The time setting scheme corresponding to the target reference vector B is that the lighting duration of the red light is 40 seconds and the lighting duration of the green light is 20 seconds. The time setting schemes corresponding to the target reference vector B are respectively 30 seconds for the lighting duration of the red light and 30 seconds for the lighting duration of the green light. The management platform can average the lighting durations corresponding to the same indicator lights in the above two time setting schemes. The final time setting scheme can be 35 seconds for the lighting duration of the red light and 25 seconds for the lighting duration of the green light.

[0137] In some embodiments, when multiple target reference vectors are determined, the management platform may also determine the time setting scheme corresponding to the highest pedestrian passing rate as the final time setting scheme among the multiple time setting schemes corresponding to the multiple target reference vectors. Figure 2And its related description. As an example only, the management platform determines the target reference vector D and the target reference vector E. The pedestrian passing rate of the time setting scheme corresponding to the target reference vector D is 89%. The pedestrian passing rate of the time setting scheme corresponding to the target reference vector E is 93%. Then the management platform can determine that the final time setting scheme is the time setting scheme corresponding to the target reference vector E.

[0138] In some embodiments of the present specification, a more reasonable time setting scheme may be determined by screening the target reference vector in combination with the pedestrian passing rate.

[0139] In some embodiments of the present specification, the management platform can use historical data as a reference, combined with historical actual conditions, to determine a more appropriate time setting plan through vector matching, thereby avoiding congestion at intersections.

[0140] The embodiments of this specification also provide a computer-readable storage medium, which stores computer instructions. When the computer reads the computer instructions in the storage medium, the computer runs any one of the smart city traffic light time setting methods provided in the embodiments of this specification.

[0141] The basic concepts have been described above. Obviously, for those skilled in the art, the above detailed disclosure is only for example and does not constitute a limitation of this specification. Although not explicitly stated here, those skilled in the art may make various modifications, improvements and corrections to this specification. Such modifications, improvements and corrections are suggested in this specification, so such modifications, improvements and corrections still belong to the spirit and scope of the exemplary embodiments of this specification.

[0142] At the same time, this specification uses specific words to describe the embodiments of this specification. For example, "one embodiment", "an embodiment", and / or "some embodiments" refer to a certain feature, structure or characteristic related to at least one embodiment of this specification. Therefore, it should be emphasized and noted that "one embodiment" or "an embodiment" or "an alternative embodiment" mentioned twice or more in different positions in this specification does not necessarily refer to the same embodiment. In addition, certain features, structures or characteristics in one or more embodiments of this specification can be appropriately combined.

[0143] In addition, unless explicitly stated in the claims, the order of the processing elements and sequences, the use of alphanumeric characters, or the use of other names described in this specification are not intended to limit the order of the processes and methods of this specification. Although the above disclosure discusses some invention embodiments that are currently considered useful through various examples, it should be understood that such details are only for illustrative purposes, and the attached claims are not limited to the disclosed embodiments. On the contrary, the claims are intended to cover all modifications and equivalent combinations that are consistent with the essence and scope of the embodiments of this specification. For example, although the management platform components described above can be implemented by hardware devices, they can also be implemented only by software solutions, such as installing the described management platform on an existing server or mobile device.

[0144] Similarly, it should be noted that in order to simplify the description disclosed in this specification and thus help understand one or more embodiments of the invention, in the above description of the embodiments of this specification, multiple features are sometimes combined into one embodiment, figure or description thereof. However, this disclosure method does not mean that the features required by the subject matter of this specification are more than the features mentioned in the claims. In fact, the features of the embodiments are less than all the features of the single embodiment disclosed above.

[0145] In some embodiments, numbers describing the number of components and attributes are used. It should be understood that such numbers used in the description of the embodiments are modified by the modifiers "about", "approximately" or "substantially" in some examples. Unless otherwise specified, "about", "approximately" or "substantially" indicate that the numbers are allowed to vary by ±20%. Accordingly, in some embodiments, the numerical parameters used in the specification and claims are approximate values, which may change according to the required features of individual embodiments. In some embodiments, the numerical parameters should take into account the specified significant digits and adopt the general method of retaining digits. Although the numerical domains and parameters used to confirm the breadth of their range in some embodiments of this specification are approximate values, in specific embodiments, the setting of such numerical values ​​is as accurate as possible within the feasible range.

[0146] Each patent, patent application, patent application publication, and other materials, such as articles, books, specifications, publications, documents, etc., cited in this specification are hereby incorporated by reference in their entirety. Except for application history documents that are inconsistent with or conflicting with the contents of this specification, documents that limit the broadest scope of the claims of this specification (currently or later attached to this specification) are also excluded. It should be noted that if the descriptions, definitions, and / or use of terms in the materials attached to this specification are inconsistent or conflicting with the contents described in this specification, the descriptions, definitions, and / or use of terms in this specification shall prevail.

[0147] Finally, it should be understood that the embodiments described in this specification are only used to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, as an example and not a limitation, the alternative configurations of the embodiments of this specification can be regarded as teachings of this specification.

Claims

1. A method for setting the time of a smart city traffic light, the method being implemented by a smart city traffic light time setting Internet of Things system, the Internet of Things system comprising a user platform, a service platform, a management platform, a sensor network platform and an object platform, the method being executed by the management platform, the method comprising: Acquiring an initial picture and intersection information of a target intersection based on the object platform through the sensor network platform, wherein the target intersection is an intersection with a traffic light, and the initial picture is a continuous picture directly collected by an image acquisition device configured by the object platform; Screening multiple frames of images in the initial picture to determine a target picture, wherein the target picture includes at least one frame of real-time image and multiple frames of historical images; The filtering of the multiple frames of images in the initial picture includes shortening the time interval of filtering two adjacent target frames in the initial picture when filtering the multiple frames of images acquired at the time point when the pedestrian approaching button is pressed; Determine pedestrian information at the target intersection based on the image recognition result of the target image, wherein the pedestrian information includes the current number of pedestrians and the number of pedestrians at multiple time points in the future; Determining the current number of pedestrians includes: Based on the at least one frame of real-time image, determining a first number of pedestrians to be determined by a pedestrian recognition model, wherein the pedestrian recognition model is a machine learning model; Based on the multiple frames of historical images, determining a second number of pedestrians to be determined by a first pedestrian prediction model, where the first pedestrian prediction model is a machine learning model; Based on the first number of pedestrians to be determined and the second number of pedestrians to be determined, the current number of pedestrians is determined by weighting, wherein the weight of the first number of pedestrians to be determined is positively correlated to the clarity of the at least one frame of real-time image, and the weight of the second number of pedestrians to be determined is positively correlated to the clarity of the multiple frames of historical images; During a time period after the pedestrian button is pressed, the weight of the first number of pedestrians to be determined is greater than the weight of the second number of pedestrians to be determined; Determining the number of pedestrians at the future multiple time points includes: Based on the multiple frames of historical images, the intersection information, event information, weather information and target time period, determining the number of pedestrians at the multiple future time points by a second pedestrian prediction model, wherein the second pedestrian prediction model is a machine learning model; The second pedestrian prediction model includes an intersection information embedding layer, an event information embedding layer and a prediction layer; The intersection information embedding layer, the event information embedding layer and the prediction layer are obtained by joint training, the training samples of the joint training include sample images, the sample images include images before the time point when the pedestrian button is pressed, and the labels of the training samples include the number of pedestrians at the time point when the pedestrian button is pressed and the time after that; Determining a time setting scheme for the traffic light at the target intersection based on the pedestrian information and the intersection information; Determining the time setting scheme includes: Based on the pedestrian information, the intersection information, the weather information and the event information, the time setting scheme is determined by vector matching; wherein, when multiple time setting schemes satisfying the vector matching conditions are matched, the one with the highest corresponding pedestrian passing rate is determined as the final time setting scheme; Determining an update frequency of the time setting scheme based on a change in a pedestrian passing rate or a pedestrian number of the time setting scheme; and The control instruction corresponding to the time setting scheme is sent to the object platform through the sensor network platform, and the object platform responds to the received control instruction to control the lighting duration of the traffic light.

2. The time setting method of a smart city traffic signal light as claimed in claim 1, characterized in that: The management platform includes several management sub-platforms.

3. The time setting method of the smart city traffic signal light as claimed in claim 2, characterized in that: The sensor network platform includes a plurality of sensor network sub-platforms; each of the plurality of sensor network sub-platforms corresponds one-to-one to each of the plurality of management sub-platforms.

4. The time setting method of a smart city traffic signal light as claimed in claim 1, characterized in that: The object platform includes a pedestrian button arranged near a crosswalk, and the pedestrian button is used to adjust the lighting duration of the traffic light.

5. A time setting Internet of Things system for traffic lights in a smart city, characterized in that: The smart city traffic light time setting IoT system includes a user platform, a service platform, a management platform, a sensor network platform and an object platform. The management platform is configured to perform the following operations: Acquiring an initial picture and intersection information of a target intersection based on the object platform through the sensor network platform, wherein the target intersection is an intersection with a traffic light, and the initial picture is a continuous picture directly collected by an image acquisition device configured by the object platform; Screening multiple frames of images in the initial picture to determine a target picture, wherein the target picture includes at least one frame of real-time image and multiple frames of historical images; The filtering of the multiple frames of images in the initial picture includes shortening the time interval of filtering two adjacent target frames in the initial picture when filtering the multiple frames of images acquired at the time point when the pedestrian approaching button is pressed; Determine pedestrian information at the target intersection based on the image recognition result of the target image, wherein the pedestrian information includes the current number of pedestrians and the number of pedestrians at multiple time points in the future; Determining the current number of pedestrians includes: Based on the at least one frame of real-time image, determining a first number of pedestrians to be determined by a pedestrian recognition model, wherein the pedestrian recognition model is a machine learning model; Based on the multiple frames of historical images, determining a second number of pedestrians to be determined by a first pedestrian prediction model, where the first pedestrian prediction model is a machine learning model; Based on the first number of pedestrians to be determined and the second number of pedestrians to be determined, the current number of pedestrians is determined by weighting, wherein the weight of the first number of pedestrians to be determined is positively correlated to the clarity of the at least one frame of real-time image, and the weight of the second number of pedestrians to be determined is positively correlated to the clarity of the multiple frames of historical images; During a time period after the pedestrian button is pressed, the weight of the first number of pedestrians to be determined is greater than the weight of the second number of pedestrians to be determined; Determining the number of pedestrians at the future multiple time points includes: Based on the multiple frames of historical images, the intersection information, event information, weather information and target time period, determining the number of pedestrians at the multiple future time points by a second pedestrian prediction model, wherein the second pedestrian prediction model is a machine learning model; The second pedestrian prediction model includes an intersection information embedding layer, an event information embedding layer and a prediction layer; The intersection information embedding layer, the event information embedding layer and the prediction layer are obtained by joint training, the training samples of the joint training include sample images, the sample images include images before the time point when the pedestrian button is pressed, and the labels of the training samples include the number of pedestrians at the time point when the pedestrian button is pressed and the time after that; Determining a time setting scheme for the traffic light at the target intersection based on the pedestrian information and the intersection information; Determining the time setting scheme includes: Based on the pedestrian information, the intersection information, the weather information and the event information, the time setting scheme is determined by vector matching; wherein, when multiple time setting schemes satisfying the vector matching conditions are matched, the one with the highest corresponding pedestrian passing rate is determined as the final time setting scheme; Determining an update frequency of the time setting scheme based on a change in a pedestrian passing rate or a pedestrian number of the time setting scheme; and The control instruction corresponding to the time setting scheme is sent to the object platform through the sensor network platform, and the object platform responds to the received control instruction to control the lighting duration of the traffic light.

6. The time setting Internet of Things system for smart city traffic lights as claimed in claim 5, characterized in that: The management platform includes several management sub-platforms.

7. The time setting Internet of Things system for smart city traffic lights as claimed in claim 6, characterized in that: The sensor network platform includes a plurality of sensor network sub-platforms; each of the plurality of sensor network sub-platforms corresponds one-to-one to each of the plurality of management sub-platforms.

8. The time setting Internet of Things system for smart city traffic lights as claimed in claim 6, characterized in that: The object platform includes a pedestrian button arranged near a crosswalk, and the pedestrian button is used to adjust the lighting duration of the traffic light.

Images