Intelligent environmental sanitation vehicle collaborative scheduling method and system based on cross-modal data

Abstract

The invention relates to the technical field of scheduling optimization, in particular to an intelligent environmental sanitation vehicle collaborative scheduling method and system based on cross-modal data, and the method comprises the steps: scheduling a first intelligent vehicle to periodically sweep a road, and collecting a falling object data set; monitoring a cleaning area of the first intelligent vehicle to obtain interference parameters; analyzing the fallen object data set through the interference parameter to obtain a reliability weight of data in the fallen object data set; a green belt anomaly recognition model is constructed to recognize the falling object data set and the reliability weight, and an anomaly region and an anomaly type are obtained; and scheduling the second smart vehicle according to the abnormal region and the abnormal type, and optimizing the control parameters. According to the invention, through the acquisition and identification of the cross-modal data, intelligent scheduling among different intelligent environmental sanitation vehicles is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of scheduling optimization, and specifically to an intelligent sanitation vehicle collaborative scheduling method and system based on cross-modal data. Background Art

[0002] The health status monitoring of urban green belts is a difficult point in urban fine management. The traditional method relying on manual inspections has high costs, low efficiency, and it is difficult to cover vast urban areas.

[0003] In recent years, the development of technologies such as the Internet of Things, big data, and artificial intelligence has provided new possibilities for improving the intelligent level of urban management. By using technologies that monitor the growth status of vegetation through remote sensing or fixed sensors, combined with intelligent camera systems for image recognition, large areas of urban green belts can be monitored and identified to improve efficiency; however, such methods have the following deficiencies: the data dimension is single, and it is easily affected by occlusion, lighting, etc.; it is difficult to accurately judge the plant growth status, pest damage, or water shortage; at the same time, with the increase in urban greening area, the deployment cost of the above methods has gradually increased.

[0004] Therefore, an intelligent sanitation vehicle collaborative scheduling method and system based on cross-modal data are proposed. Summary of the Invention

[0005] The purpose of the present invention is to provide an intelligent sanitation vehicle collaborative scheduling method and system based on cross-modal data; including: scheduling a first intelligent vehicle to periodically clean the road, and collecting a fallen object data set; monitoring the cleaning area of the first intelligent vehicle to obtain interference parameters; analyzing the fallen object data set through the interference parameters to obtain the reliability weights of the data in the fallen object data set; constructing a green belt anomaly recognition model to identify the fallen object data set and the reliability weights to obtain the anomaly area and anomaly type; scheduling a second intelligent vehicle according to the anomaly area and anomaly type, and optimizing the control parameters; through the identification of the anomaly area and anomaly type, realizing intelligent scheduling between different intelligent sanitation vehicles.

[0006] To achieve the above purpose, the present invention provides the following technical solutions: An intelligent sanitation vehicle collaborative scheduling method based on cross-modal data, including: Scheduling a first intelligent vehicle to periodically clean the road, and collecting a fallen object data set; the fallen object data set includes fallen object parameters, fallen object types, fallen object weights, and fallen object moisture contents; Monitoring the cleaning area of the first intelligent vehicle to obtain interference parameters; the interference parameters include environmental interference parameters, vehicle interference parameters, and terrain interference parameters; Analyze the dropped object dataset through interference parameters to obtain the reliability weights of the data in the dropped object dataset; Construct a green belt anomaly recognition model to recognize the dropped object dataset and reliability weights, and obtain the anomaly areas and anomaly types; Dispatch the second intelligent vehicle according to the anomaly areas and anomaly types, and optimize the control parameters.

[0007] The process of obtaining the dropped object dataset includes: Set a camera device on the first intelligent vehicle to recognize the road surface and the in-vehicle storage bin, and obtain the dropped object parameters; the dropped object parameters are dropped object images; the first intelligent vehicle is an intelligent sweeper; Recognize the dropped object images to obtain the dropped object types; the dropped object types include vegetation leaves, vegetation branches, vegetation fruits, vegetation flowers, and green belt soil; Set a weighing device in the first intelligent vehicle, and combine with the recognition of the dropped object images to obtain the dropped object weights; Set a moisture content detection device in the first intelligent vehicle to detect the moisture content of the dropped objects.

[0008] The process of obtaining the interference parameters includes: Monitor the cleaning area of the first intelligent vehicle through Internet of Things sensors to obtain environmental interference parameters; the environmental interference parameters include environmental wind direction, environmental wind speed, environmental temperature, and environmental humidity; Monitor the passing vehicles in the road area through the traffic camera device of the road to obtain vehicle interference parameters; the vehicle interference parameters include vehicle shape and vehicle speed; Recognize the terrain of the cleaning area of the first intelligent vehicle to obtain terrain interference parameters, and the terrain interference parameters include road slope.

[0009] The process of recognizing the reliability weights includes: Obtain the dropped object dataset and interference parameters; Recognize according to the dropped object parameters in the dropped object dataset to obtain the dropped object shapes; Recognize according to the dropped object shapes, dropped object weights, terrain interference parameters, vehicle interference parameters, and environmental interference parameters to obtain displacement prediction values of different dropped objects; recognize according to the displacement prediction values to obtain distribution reliability coefficients; Normalize the distribution reliability coefficients to obtain reliability weights.

[0010] The green belt anomaly recognition model includes an anomaly area judgment layer and an anomaly type recognition layer; The abnormal area judgment layer identifies the dropped object data set and the reliability weight; according to the data characteristics and reliability weight of each data dimension in the dropped object data set, it identifies the area abnormal coefficient; and makes a judgment based on the area abnormal coefficient to determine the abnormal area; the data characteristics include the time characteristics and space characteristics of the data in the dropped object data set. The abnormal type identification layer obtains the dropped object data set and the reliability weight of the abnormal area, and identifies the abnormal type of the abnormal area, where the abnormal types include drought, disease, and pest damage.

[0011] The optimization process of the control parameters includes: When there is an abnormal area in the green belt, dispatch the second intelligent vehicle; and generate the loading parameters, trajectory parameters, and spraying parameters of the second intelligent vehicle according to the location, abnormal type, and area abnormal coefficient of the abnormal area, and perform parameter optimization. The second intelligent vehicle is an intelligent spraying vehicle; the loading parameters are the liquid parameters loaded by the second intelligent vehicle; the trajectory parameters are the driving trajectory parameters of the second intelligent vehicle; the spraying parameters are the spraying control parameters of the second intelligent vehicle in the abnormal area.

[0012] A smart environmental sanitation vehicle collaborative scheduling system based on cross-modal data includes: The dropped object data acquisition module dispatches the first intelligent vehicle to periodically clean the road and collects the dropped object data set; the dropped object data set includes dropped object parameters, dropped object types, dropped object weights, and dropped object moisture contents. The interference parameter acquisition module monitors the cleaning area of the first intelligent vehicle to obtain interference parameters; the interference parameters include environmental interference parameters, vehicle interference parameters, and terrain interference parameters. The reliability identification module analyzes the dropped object data set through the interference parameters to obtain the reliability weight of the data in the dropped object data set. The abnormal identification module constructs a green belt abnormal identification model to identify the dropped object data set and the reliability weight, and obtains the abnormal area and abnormal type. The scheduling optimization module dispatches the second intelligent vehicle according to the abnormal area and abnormal type, and performs optimization of the control parameters.

[0013] Compared with the prior art, the beneficial effects of the present invention are: 1. The present invention schedules the first intelligent vehicle to clean the road. Through the intelligent device installed on the first intelligent vehicle, the collected waste is identified to obtain a fallen object data set, including fallen object parameters, fallen object types, fallen object weights, fallen object moisture contents, etc., which can accurately obtain data reflecting the green belt vegetation. And by combining the data acquisition method with the cleaning vehicle of intelligent environmental sanitation, the cost of separately setting data acquisition equipment can be greatly reduced.

[0014] 2. According to the comprehensive identification of the characteristics of the fallen object itself, the influence of wind force, and the influence of terrain, the present application predicts the displacement prediction values of fallen objects of different vegetation, and identifies the distribution reliability coefficient according to the displacement prediction values and displacement thresholds. Through normalization processing of the distribution reliability coefficient, the reliability weight is obtained. The reliability of the data in the fallen object data set is accurately measured through the reliability weight, providing a data basis for the accurate identification of green belt anomalies.

[0015] 3. The present invention constructs a green belt anomaly identification model to identify the fallen object data set and the reliability weight. According to the time characteristics, spatial characteristics of each data dimension in the fallen object data set and the corresponding reliability weight, the anomaly coefficient is identified to determine the anomaly area. The fallen object data set and the reliability weight of the anomaly area are obtained to identify the anomaly type of the anomaly area. Thus, the anomaly area and anomaly type of the green belt are accurately identified. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic flow chart of a method for collaborative scheduling of intelligent environmental sanitation vehicles based on cross-modal data according to the present invention; Figure 2 It is a schematic structural diagram of a green belt anomaly identification model according to the present invention; Figure 3 It is a schematic diagram of the collaborative scheduling logic between intelligent environmental sanitation vehicles according to the present invention; Figure 4 It is a schematic structural diagram of a system for collaborative scheduling of intelligent environmental sanitation vehicles based on cross-modal data according to the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0017] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0018] The health status monitoring of urban green belts is a difficult point in the refined management of cities. With the development of technologies such as the Internet of Things, big data, and artificial intelligence, through the technology of using remote sensing or fixed sensors to monitor the growth status of vegetation, combined with an intelligent camera system for image recognition, large-area urban green belts can be monitored and identified to improve efficiency. However, the data dimension of the above methods is single, and the data interference is serious, making it difficult to accurately identify. At the same time, the use and deployment of hardware result in high costs.

[0019] At the same time, with the development of the intelligent level of urban management in intelligent environmental sanitation, certain results have been achieved in improving the efficiency of garbage collection and transportation and optimizing resource allocation. Intelligent environmental sanitation mainly focuses on operation management and resource scheduling optimization, regarding the waste collected by cleaning as a simple treatment object, and failing to fully explore the potential information value contained in these wastes themselves. Therefore, by combining the monitoring of urban green belts with intelligent environmental sanitation, during the process of intelligent sanitation vehicles maintaining urban environmental hygiene, the garbage and waste cleaned are identified, and the vegetation of urban green belts is monitored. And a collaborative scheduling method and system for intelligent sanitation vehicles based on cross-modal data are proposed.

[0020] Embodiment 1 The present invention proposes a collaborative scheduling method for intelligent sanitation vehicles based on cross-modal data. Its process is shown in Figure 1 and includes: Scheduling the first intelligent vehicle to periodically clean the road and collecting a dropped object data set. The dropped object data set includes dropped object parameters, dropped object types, dropped object weights, and dropped object moisture contents.

[0021] During the acquisition process of the dropped object data set, the road is divided into stages as data collection points. Since the road width at the green belt is generally constructed according to a fixed standard, it is divided according to the length of the road, including but not limited to, taking a 10-meter-long road as a data collection point. For the waste collected by the intelligent cleaning vehicle within this data collection point, it is identified to obtain the dropped object data set.

[0022] The acquisition process of the dropped object data set includes: Setting a camera device on the first intelligent vehicle to identify the road surface and the storage bin in the vehicle to obtain dropped object parameters. The dropped object parameters are dropped object images. The first intelligent vehicle is an intelligent cleaning vehicle. The intelligent cleaning vehicle includes a suction-sweeping road sweeper, and its working principle includes: using a rotary brush to sweep the waste on the road surface to the ground in the middle of the vehicle, and then sucking the waste into the storage bin in the vehicle through a vacuum suction nozzle. The storage bin is used to store the waste cleaned by the intelligent cleaning vehicle.

[0023] Identify the images of the fallen objects to obtain the types of fallen objects; the types of fallen objects include vegetation leaves, vegetation branches, vegetation fruits, vegetation flowers, and green belt soil; Install a weighing device in the first intelligent vehicle. Combining the identification of the fallen object images, obtain the weight of the fallen objects. Since there are many wastes swept on the first intelligent vehicle and it is difficult to accurately weigh the fallen objects, weight prediction is performed by combining image recognition, and the weight data of the weighing device is corrected and calibrated to obtain the fallen object weight data; Install a moisture content detection device in the first intelligent vehicle to detect the moisture content of the fallen objects. Commonly used moisture content detection devices include: near-infrared sensors, microwave moisture sensors, capacitive moisture sensors, etc.; among them, the near-infrared sensor detects the reflectivity of the object in the near-infrared band. Different moisture contents cause changes in the reflection characteristics, so as to deduce the moisture content inversely; it can perform non-contact measurement, is suitable for dynamic operations, and has a fast response speed and can scan a large area.

[0024] Taking the camphor tree as an example, the camphor tree is a common greening tree species; the camphor tree is an evergreen tree species, but its leaves do not fall permanently. Instead, it completes metabolism by concentrating leaf replacement in spring. Its fallen objects mainly include fallen leaves, branches, and fruits. After the turn of spring and summer, clean the road near the camphor tree green belt to obtain the fallen object data as shown in Table 1.

[0025] Table 1 Fallen Object Data Table

[0026] It should be noted that the division of vegetation leaves, vegetation branches, vegetation fruits, vegetation flowers, and green belt soil in the types of fallen objects is only a relatively broad division by type. In the actual implementation process, it can be refined according to the characteristics of the fallen objects to improve the accuracy of cross-modal data. For example, for vegetation leaves, they can be divided into green leaves, yellow leaves, brown leaves, etc. according to the color of the leaves, and can also be distinguished according to the types of vegetation; vegetation branches can be divided according to the length of the branches, whether there are branches and leaves, etc.

[0027] The division of the types of fallen objects can be verified multiple times according to the collected historical data, and finally determine the combination of type divisions with the best recognition efficiency and accuracy.

[0028] The present invention schedules the first intelligent vehicle to clean the road. Through the intelligent devices installed on the first intelligent vehicle, the collected wastes are identified to obtain a fallen object data set, including fallen object parameters, fallen object types, fallen object weights, and fallen object moisture contents, etc., and can accurately obtain data reflecting the green belt vegetation; and combining the data acquisition method with the cleaning vehicle of intelligent environmental sanitation can greatly reduce the cost of separately setting data acquisition devices.

[0029] Monitor the cleaning area of the first intelligent vehicle to obtain interference parameters; the interference parameters include environmental interference parameters, vehicle interference parameters, and terrain interference parameters.

[0030] The process of obtaining the interference parameters includes: Monitor the cleaning area of the first intelligent vehicle through Internet of Things sensors to obtain environmental interference parameters; the environmental interference parameters include environmental wind direction, environmental wind speed, environmental temperature, and environmental humidity; Monitor the passing vehicles in the road area through the traffic camera device of the road to obtain vehicle interference parameters; the vehicle interference parameters include vehicle shape and vehicle speed; the vehicle shape can be directly obtained through image recognition, or the vehicle model can be recognized, and the vehicle shape data can be obtained according to the vehicle model and the preset vehicle database.

[0031] Identify the terrain of the cleaning area of the first intelligent vehicle to obtain terrain interference parameters, and the terrain interference parameters include road slope.

[0032] The present invention monitors the cleaning area of the first intelligent vehicle through Internet of Things sensors to obtain environmental interference parameters; monitors the passing vehicles in the road area through the traffic camera device of the road to obtain vehicle interference parameters; identifies the terrain of the cleaning area of the first intelligent vehicle to obtain terrain interference parameters; thereby accurately identifying the interference parameters and providing a data basis for the reliable identification of data in the dropped object dataset.

[0033] Analyze the dropped object dataset through the interference parameters to obtain the reliability weight of the data in the dropped object dataset.

[0034] Due to the small weight of the dropped objects of vegetation and the soil of the green belt, the natural wind in the environment and the driving wind generated during the vehicle driving may cause the dropped objects to move; once the dropped objects move, it will cause errors in the data obtained at different data collection points, thereby affecting the accuracy of the abnormal identification of the green belt.

[0035] Therefore, this application proposes a reliability weight, starting from the displacement of the dropped objects to judge the reliability of its data; the process of identifying the reliability weight includes: Obtain the dropped object dataset and interference parameters; Identify according to the dropped object parameters in the dropped object dataset to obtain the dropped object shape; Identify according to the dropped object shape, dropped object weight, terrain interference parameters, vehicle interference parameters, and environmental interference parameters to obtain the displacement prediction values of different dropped objects; identify according to the displacement prediction values to obtain the distribution reliability coefficient; Normalize the distribution reliability coefficient to obtain the reliability weight.

[0036] In the process of obtaining the reliability weight, the present application mainly considers three aspects of factors, including the characteristics of the falling objects themselves, the influence of wind force, and the influence of terrain. The characteristics of the falling objects themselves include the shape and weight of the falling objects. Shape of the falling objects: For vegetation leaves, leaves with a larger area and a broad shape are more likely to be affected by wind force, while leaves that are thicker and rougher are not easily blown by the wind. For vegetation fruits and green belt soil, the closer the shape is to a sphere, the easier it is to move. Weight of the falling objects: The heavier the falling object, the less likely it is to move, and the lighter the falling object, the easier it is to move.

[0037] Wind force: The greater the wind force, the easier it is for the falling object to move; the smaller the wind force, the less likely it is for the falling object to move.

[0038] Terrain: Determine the moving direction of the falling object according to the terrain distribution of the road and the wind direction, and then determine the road slope in the moving direction. The smaller the road slope, the easier it is for the falling object to move; the greater the road slope, the less likely it is for the falling object to move.

[0039] The present application predicts the displacement prediction values of different vegetation falling objects through the comprehensive recognition of the characteristics of the falling objects themselves, the influence of wind force, and the influence of terrain, and obtains the distribution reliability coefficient according to the displacement prediction value and the displacement threshold; performs normalization processing according to the distribution reliability coefficient to obtain the reliability weight; accurately measures the data reliability in the falling object dataset through the reliability weight, providing a data basis for the accurate recognition of green belt anomalies.

[0040] In particular, in order to further improve the recognition of green belt anomalies, the data in the falling object dataset can also be re-recognized and calibrated, including: Calibrate the moisture content of the falling objects according to the environmental temperature and environmental humidity in the environmental interference parameters. Calibrate the reliability of the data of the falling object types that do not exist at the current data collection point according to the falling object types. That is, if the falling object type matches the green vegetation at the data collection point, the data reliability is relatively high; if there are falling object types that do not belong to the green vegetation at the data collection point, it is considered that this part of the falling object types has moved from other areas, then delete the relevant data of this part of the falling object types, and correct the reliability according to the proportion of its weight or volume.

[0041] For example, only vegetation A exists in data collection point 1; when collecting falling object data, the falling object types include the leaves of vegetation A and the leaves of vegetation B, then delete the relevant data of the leaves of vegetation B, perform reliability analysis according to the remaining falling object data, and re-adjust the reliability weight of the data of the leaves of vegetation A according to the weight ratio of the leaves of vegetation A and the leaves of vegetation B, and the adjustment range is verified and determined according to the corresponding historical data.

[0042] Construct a green belt anomaly recognition model to recognize the dropped object dataset and reliability weights, and obtain the anomaly area and anomaly type.

[0043] The structure of the green belt anomaly recognition model is as Figure 2 shown, including an anomaly area judgment layer and an anomaly type recognition layer; The anomaly area judgment layer recognizes the dropped object dataset and reliability weights; according to the data characteristics and reliability weights of each data dimension in the dropped object dataset, the area anomaly coefficient is recognized; according to the area anomaly coefficient, a judgment is made to determine the anomaly area; the data characteristics include the time characteristics and spatial characteristics of the data in the dropped object dataset; The anomaly type recognition layer obtains the dropped object dataset and reliability weights of the anomaly area, and recognizes the anomaly type of the anomaly area. The anomaly types include drought, disease, and pest.

[0044] The present invention constructs a green belt anomaly recognition model to recognize the dropped object dataset and reliability weights; according to the time characteristics, spatial characteristics and corresponding reliability weights of each data dimension in the dropped object dataset, the anomaly coefficient is recognized, and the anomaly area is determined; the dropped object dataset and reliability weights of the anomaly area are obtained, and the anomaly type of the anomaly area is recognized; thus, the anomaly area and anomaly type of the green belt are accurately recognized.

[0045] Dispatch the second intelligent vehicle according to the anomaly area and anomaly type, and optimize the control parameters. The collaborative dispatch logic among intelligent sanitation vehicles is as Figure 3 shown.

[0046] The optimization process of the control parameters includes: When there is an anomaly area in the green belt, dispatch the second intelligent vehicle; and according to the location, anomaly type and area anomaly coefficient of the anomaly area, generate the loading parameters, trajectory parameters and spraying parameters of the second intelligent vehicle, and optimize the parameters; The second intelligent vehicle is an intelligent spraying vehicle; the loading parameters are the liquid parameters loaded by the second intelligent vehicle; the trajectory parameters are the driving trajectory parameters of the second intelligent vehicle; the spraying parameters are the spraying control parameters of the second intelligent vehicle in the anomaly area. The liquid parameters include water for drought, water for pests, and water for diseases; the spraying control parameters include spraying flow rate.

[0047] The present invention recognizes the anomaly area and anomaly type of the green belt, dispatches the second intelligent vehicle to process the area with anomalies; and generates and optimizes the loading parameters, trajectory parameters and spraying parameters according to the anomaly type and area anomaly coefficient, improving the processing efficiency of the anomaly area of the green belt.

[0048] The present invention also provides a collaborative scheduling system for intelligent environmental sanitation vehicles based on cross-modal data, and its structure is as Figure 4 shown; including: a falling object data acquisition module, which schedules the first intelligent vehicle to periodically clean the road and collects a falling object data set; the falling object data set includes falling object parameters, falling object types, falling object weights, and falling object moisture contents; An interference parameter acquisition module monitors the cleaning area of the first intelligent vehicle to obtain interference parameters; the interference parameters include environmental interference parameters, vehicle interference parameters, and terrain interference parameters; A reliability identification module analyzes the falling object data set through the interference parameters to obtain the reliability weights of the data in the falling object data set; An anomaly identification module constructs a green belt anomaly identification model to identify the falling object data set and the reliability weights, and obtains the anomaly area and anomaly type; A scheduling optimization module schedules the second intelligent vehicle according to the anomaly area and anomaly type, and optimizes the control parameters.

[0049] The present invention schedules the first intelligent vehicle to periodically clean the road and collects a falling object data set; monitors the cleaning area of the first intelligent vehicle to obtain interference parameters; analyzes the falling object data set through the interference parameters to obtain the reliability weights of the data in the falling object data set; constructs a green belt anomaly identification model to identify the falling object data set and the reliability weights, and obtains the anomaly area and anomaly type; schedules the second intelligent vehicle according to the anomaly area and anomaly type, and optimizes the control parameters; through the acquisition and identification of cross-modal data, realizes the intelligent scheduling between different intelligent environmental sanitation vehicles.

[0050] Embodiment 2 The present invention provides a collaborative scheduling method for intelligent environmental sanitation vehicles based on cross-modal data, including: Scheduling the first intelligent vehicle to periodically clean the road and collecting a falling object data set; the falling object data set includes falling object parameters, falling object types, falling object weights, and falling object moisture contents.

[0051] The acquisition process of the falling object data set includes: Setting a camera device on the first intelligent vehicle to identify the road surface and the in-vehicle storage bin, and obtaining the falling object parameters; the falling object parameters are falling object images; the first intelligent vehicle is an intelligent sweeper; Identifying the falling object images to obtain the falling object types; the falling object types include vegetation leaves, vegetation branches, vegetation fruits, vegetation flowers, and green belt soil; Setting a weighing device in the first intelligent vehicle and combining the identification of the falling object images to obtain the falling object weights; A water content detection device is arranged inside the first intelligent vehicle to detect the water content of the fallen objects.

[0052] The present invention schedules the first intelligent vehicle to clean the road. Through the intelligent device arranged on the first intelligent vehicle, the collected waste is identified to obtain a fallen object data set, including fallen object parameters, fallen object types, fallen object weights, and fallen object water contents, etc., and data reflecting the green belt vegetation can be accurately obtained; and by combining the data acquisition method with the cleaning vehicle of intelligent environmental sanitation, the cost of separately setting data acquisition equipment can be greatly reduced.

[0053] The cleaning area of the first intelligent vehicle is monitored to obtain interference parameters; the interference parameters include environmental interference parameters, vehicle interference parameters, and terrain interference parameters.

[0054] The process of obtaining the interference parameters includes: The cleaning area of the first intelligent vehicle is monitored through an Internet of Things sensor to obtain environmental interference parameters; the environmental interference parameters include environmental wind direction, environmental wind speed, environmental temperature, and environmental humidity; The passing vehicles in the road area are monitored through the traffic camera device of the road to obtain vehicle interference parameters; the vehicle interference parameters include vehicle shape and vehicle speed; The terrain of the cleaning area of the first intelligent vehicle is identified to obtain terrain interference parameters, and the terrain interference parameters include road slope.

[0055] The present invention monitors the cleaning area of the first intelligent vehicle through an Internet of Things sensor to obtain environmental interference parameters; monitors the passing vehicles in the road area through the traffic camera device of the road to obtain vehicle interference parameters; identifies the terrain of the cleaning area of the first intelligent vehicle to obtain terrain interference parameters; thereby accurately identifying the interference parameters and providing a data basis for the reliable identification of the data in the fallen object data set.

[0056] The fallen object data set is analyzed through the interference parameters to obtain the reliability weight of the data in the fallen object data set.

[0057] The process of identifying the reliability weight includes: Obtain the fallen object data set and the interference parameters; Identify according to the fallen object parameters in the fallen object data set to obtain the fallen object shape; Identify according to the fallen object shape, fallen object weight, terrain interference parameters, vehicle interference parameters, and environmental interference parameters to obtain the displacement prediction values of different fallen objects; identify according to the displacement prediction values to obtain the distribution reliability coefficient; Normalize the distribution reliability coefficient to obtain the reliability weight.

[0058] Based on the comprehensive identification of the characteristics of the falling objects themselves, the influence of wind, and the influence of terrain, this application predicts the displacement prediction values of different vegetation falling objects, and identifies the distribution reliability coefficient according to the displacement prediction values and the displacement threshold; performs normalization processing according to the distribution reliability coefficient to obtain the reliability weight; accurately measures the data reliability in the falling object data set through the reliability weight, providing a data basis for the accurate identification of green belt anomalies.

[0059] Construct a green belt anomaly identification model to identify the falling object data set and the reliability weight, and obtain the anomaly area and anomaly type.

[0060] The green belt anomaly identification model includes an anomaly area judgment layer and an anomaly type identification layer; The anomaly area judgment layer identifies the falling object data set and the reliability weight; according to the data characteristics and reliability weight of each data dimension in the falling object data set, identifies the area anomaly coefficient; makes a judgment according to the area anomaly coefficient to determine the anomaly area; the data characteristics include the time characteristics and spatial characteristics of the data in the falling object data set. The anomaly type identification layer obtains the falling object data set and the reliability weight of the anomaly area, and identifies the anomaly type of the anomaly area, and the anomaly types include drought, disease, and pest damage.

[0061] This invention constructs a green belt anomaly identification model to identify the falling object data set and the reliability weight; according to the time characteristics, spatial characteristics and corresponding reliability weight of each data dimension in the falling object data set, identifies the anomaly coefficient and determines the anomaly area; obtains the falling object data set and the reliability weight of the anomaly area, and identifies the anomaly type of the anomaly area; thus accurately identifying the anomaly area and anomaly type of the green belt.

[0062] Dispatch the second intelligent vehicle according to the anomaly area and anomaly type, and optimize the control parameters.

[0063] The optimization process of the control parameters includes: When there is an anomaly area in the green belt, dispatch the second intelligent vehicle; and generate the loading parameters, trajectory parameters and spraying parameters of the second intelligent vehicle according to the location, anomaly type and area anomaly coefficient of the anomaly area, and perform parameter optimization; The second intelligent vehicle is an intelligent spraying vehicle; the loading parameter is the liquid parameter loaded by the second intelligent vehicle; the trajectory parameter is the driving trajectory parameter of the second intelligent vehicle; the spraying parameter is the spraying control parameter of the second intelligent vehicle in the anomaly area.

[0064] The present invention identifies abnormal areas and abnormal types in the green belt, schedules the second intelligent vehicle to process the areas with abnormalities, and generates and optimizes loading parameters, trajectory parameters, and spraying parameters according to the abnormal types and area abnormality coefficients, so as to improve the processing efficiency of the abnormal areas in the green belt.

[0065] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A smart sanitation vehicle collaborative scheduling method based on cross-modal data, characterized in that: include: Dispatching a first intelligent vehicle to periodically clean the road to collect a dropped object data set; the dropped object data set includes dropped object parameters, dropped object types, dropped object weights, and dropped object moisture content; Monitoring the cleaning area of ​​the first intelligent vehicle to obtain interference parameters; the interference parameters include environmental interference parameters, vehicle interference parameters and terrain interference parameters; The dropped object data set is analyzed by using interference parameters to obtain the reliability weight of the data in the dropped object data set; A green belt anomaly recognition model is built to identify the dropped object data set and reliability weights, and the anomaly area and anomaly type are obtained; The second intelligent vehicle is dispatched according to the abnormal area and abnormal type, and the control parameters are optimized.

2. According to claim 1, a smart sanitation vehicle collaborative scheduling method based on cross-modal data is characterized by: The process of acquiring the dropped object dataset includes: A camera device is arranged on the first smart vehicle to identify the road surface and the storage bin in the vehicle to obtain the parameters of the dropped object; the dropped object parameters are the dropped object images; the first smart vehicle is a smart sweeper; Identify the dropped object image to obtain the type of dropped object; the dropped object types include vegetation leaves, vegetation branches, vegetation fruits, vegetation flowers and green belt soil; A weighing device is arranged in the first intelligent vehicle, and the weight of the dropped object is obtained by combining the recognition of the dropped object image; A moisture content detection device is provided in the first intelligent vehicle to detect the moisture content of the dropped object.

3. According to claim 1, a method for collaborative dispatching of intelligent sanitation vehicles based on cross-modal data is characterized in that: The process of obtaining the interference parameters includes: The cleaning area of ​​the first smart vehicle is monitored by an Internet of Things sensor to obtain environmental interference parameters; the environmental interference parameters include environmental wind direction, environmental wind speed, environmental temperature and environmental humidity; The vehicles passing through the road area are monitored by a traffic camera device on the road to obtain vehicle interference parameters; the vehicle interference parameters include vehicle shape and vehicle speed; The terrain of the cleaning area of ​​the first intelligent vehicle is identified to obtain terrain interference parameters, where the terrain interference parameters include a road slope.

4. The method for collaborative scheduling of intelligent sanitation vehicles based on cross-modal data according to claim 1 is characterized in that: The identification process of the reliability weight includes: Obtain the falling object dataset and interference parameters; Identify the dropped objects according to the dropped object parameters in the dropped object data set to obtain the shape of the dropped objects; According to the shape of the dropped object, the weight of the dropped object, the terrain interference parameter, the vehicle interference parameter and the environmental interference parameter, the displacement prediction value of different dropped objects is obtained; according to the displacement prediction value, the distribution reliability coefficient is obtained by identification; The distribution reliability coefficient is normalized to obtain a reliability weight.

5. According to claim 1, a method for collaborative scheduling of intelligent sanitation vehicles based on cross-modal data is characterized in that: The green belt anomaly recognition model includes an abnormal area judgment layer and an abnormal type recognition layer; The abnormal area judgment layer identifies the dropped object data set and the reliability weight; identifies the regional abnormality coefficient according to the data characteristics and reliability weight of each data dimension in the dropped object data set; determines the abnormal area according to the regional abnormality coefficient; the data characteristics include the time characteristics and spatial characteristics of the data in the dropped object data set; The abnormal type identification layer obtains the dropped object data set and reliability weight of the abnormal area, and identifies the abnormal type of the abnormal area, wherein the abnormal type includes drought, disease and insect pest.

6. The method for collaborative dispatching of intelligent sanitation vehicles based on cross-modal data according to claim 5 is characterized by: The optimization process of the control parameters includes: When there is an abnormal area of ​​the green belt, dispatch the second smart vehicle; and generate the loading parameters, trajectory parameters and spraying parameters of the second smart vehicle according to the location, abnormal type and regional abnormality coefficient of the abnormal area, and perform parameter optimization; The second smart vehicle is a smart spraying vehicle; the loading parameters are parameters of the liquid loaded on the second smart vehicle; the trajectory parameters are parameters of the driving trajectory of the second smart vehicle; and the spraying parameters are spraying control parameters of the second smart vehicle in the abnormal area.

7. A smart sanitation vehicle collaborative dispatching system based on cross-modal data, characterized in that: include: A dropped object data acquisition module, which dispatches the first intelligent vehicle to periodically clean the road to collect a dropped object data set; the dropped object data set includes dropped object parameters, dropped object types, dropped object weights, and dropped object moisture content; An interference parameter acquisition module monitors the cleaning area of ​​the first intelligent vehicle to obtain interference parameters; the interference parameters include environmental interference parameters, vehicle interference parameters and terrain interference parameters; The reliability identification module analyzes the dropped object data set through interference parameters to obtain the reliability weight of the data in the dropped object data set; The anomaly recognition module builds a green belt anomaly recognition model to identify the dropped object data set and reliability weights, and obtain the anomaly area and anomaly type; The scheduling optimization module schedules the second intelligent vehicle according to the abnormal area and abnormal type, and optimizes the control parameters.

Images