A 5G smart park ground garbage identification method and system

Abstract

The application discloses a 5G smart park ground garbage identification method and system, uses monitoring equipment in the park to collect ground images, thereby realizing real-time monitoring of the ground in the park and facilitating quick acquisition of garbage information; the collected images are preprocessed to ensure image quality and data accuracy; image recognition technology is used to classify and identify the images, and the types and positions of the garbage are identified; the identified data are processed, the types and quantities of the garbage are analyzed, and the data are uploaded to a cloud database, thereby facilitating further data analysis and management; according to the ground garbage information recorded in the cloud database, a route is planned, and a garbage cleaning vehicle is dispatched to the site to process the garbage. The loss function of the network calculates additional loss for the part inconsistent with the actual segmentation result, guides the network to pay attention to the part prone to misclassification, and the transfer of the cleaning task in the planning process can efficiently mobilize all cleaning vehicles in the park.

Description

Technical Field

[0001] The present invention relates to the technical field of ground garbage identification, and in particular to a ground garbage identification method and system based on 5G smart park. Background Art

[0002] As the amount of garbage generated by human activities continues to increase, traditional garbage management methods are becoming unsustainable. This leads to the need for more innovative solutions to deal with the increasing amount of garbage and ensure a cleaner and more sustainable environment. 5G technology provides the high-speed, low-latency communication necessary for real-time transmission of large amounts of data, thereby providing reliable support for the continuous monitoring and analysis of garbage within the park in the smart park ground garbage identification demand. In addition, 5G technology can also collaborate with advanced technologies such as computer vision, artificial intelligence and the Internet of Things to perform real-time processing, analysis and decision-making on image and video data within the park. The present invention proposes a 5G-based smart park ground garbage identification method and system to meet the growing demand for more efficient and effective waste management solutions, which is of great significance to promoting a cleaner and more sustainable environment. Summary of the Invention

[0003] In view of this, the present invention provides a method and system for identifying ground garbage in a 5G smart park, the purpose of which is to: (1) use image recognition technology and deep learning algorithms to achieve rapid and accurate identification and classification of garbage, thereby improving the environmental sanitation conditions of the park; (2) through the high-speed data transmission capability of the 5G network, achieve real-time monitoring and management of the garbage identification system, help optimize the resource allocation of garbage collection and treatment, and thus improve the efficiency of garbage treatment.

[0004] To achieve the above objectives, the present invention provides a method and system for identifying ground waste in a 5G smart park, comprising the following steps:

[0005] S1: Use monitoring equipment within the park to collect ground images, thereby monitoring the ground within the park in real time, making it easy to quickly obtain spam information;

[0006] S2: Preprocess the collected images to ensure image quality and data accuracy;

[0007] S3: Use image recognition technology to classify and identify images and the type and location of garbage;

[0008] S4: Process the identified data, analyze the type and quantity of garbage, and upload it to the cloud database to facilitate further data analysis and management;

[0009] S5: Based on the ground garbage information recorded in the cloud database, a route is planned to dispatch garbage collection trucks to the site to process the garbage.

[0010] Optionally, the S1 step uses the monitoring device in the park to collect ground images, thereby monitoring the ground in the park in real time, facilitating the rapid acquisition of garbage information, including:

[0011] Monitoring devices are arranged at different positions in the park to collect ground images. The collected images are transmitted to the data processing server through the 5G network for subsequent image recognition algorithm analysis. The arranged monitoring devices need to be able to monitor the entire ground of the smart park, so that the subsequent ground garbage recognition method can comprehensively identify the entire park.

[0012] Optionally, the S2 step pre-processes the collected images to ensure image quality and data accuracy, including:

[0013] The pre-processing of the image includes two parts of image denoising and image foreground extraction. The image denoising directly uses the median filtering method. The foreground extraction algorithm flow of the image includes:

[0014] S21: According to the maximum and minimum values of the image, a series of threshold values are selected to calculate the foreground and background segmentation results of multiple threshold values:

[0015] ;

[0016] ;

[0017] ;

[0018] Wherein, S represents the segmentation result image of the image foreground and background, is a pixel in the image S, b is the color channel of the image, which is one of red, green and blue, P is the ground image, is a pixel in the ground image V, (i,j) represents the horizontal and vertical coordinates of the pixel, represents the minimum value in the image, represents the maximum value in the image;

[0019] S22: Select the range of 10% of the edge of the ground image as the reference background, and compare the part coinciding with the reference background in the series of foreground and background segmentation results obtained as the similarity. The calculation method of the similarity is:

[0020] ;

[0021] Wherein, B represents the reference background point coordinate set, is the number of reference background points, the function is used to identify whether the pixel in the segmentation result is background, ;

[0022] S23: After obtaining the similarity of the foreground and background segmentation results under different thresholds and the reference background, the segmentation results with a similarity exceeding 50% are removed, and each pixel of the ground image is voted based on the remaining segmentation results. If the number of votes for the current pixel as foreground or background exceeds half of the total number of votes, the pixel is determined as foreground or background. Finally, all segmentation images are merged into a final foreground and background segmentation result of the ground image.

[0023] Optionally, in the S3 step, image recognition technology is used to classify and identify the image, and the types and positions of the garbage are identified, including:

[0024] S31: The ground image and the foreground and background segmentation result obtained in step S2 are merged to obtain a merged image as an input image of the image recognition technology:

[0025] ;

[0026] wherein, is the foreground and background segmentation result obtained in step S2;

[0027] S32: The image is input into a Yolo-based target detection deep learning network to obtain the identified types and positions of the ground garbage, which are described by the following parameters:

[0028] ;

[0029] wherein represents the pixel position of the ground image where the center of the ground garbage detection frame is located; represents the width and height of the detection frame where the detected garbage is located; represents the probability that the garbage in the detection frame belongs to different types of garbage, and N represents the number of garbage types to be identified, represents the probability that there is no garbage in the detection frame;

[0030] S33: After the network outputs the parameters of the predicted detection frame, the loss function is calculated by comparing with the actual detection frame for training the network, and the calculation method of the loss function is:

[0031] ;

[0032] ;

[0033] ;

[0034] ;

[0035] ;

[0036] wherein S represents the number of grids defined in the Yolo-based target detection deep learning network; represents 1 if there is garbage in the grid, otherwise 0; and respectively represent the center position of the detection frame where the garbage is located and the real value of the width and height of the detection frame when there is garbage in the grid and the value predicted by the network; represents 1 if the grid contains the mth type of garbage, otherwise 0; represents 1 if there is no garbage in the grid, otherwise 0; represents the probability that the garbage in the detection frame belongs to different types of garbage, and N represents the number of types of garbage to be identified, represents the probability that there is no garbage in the detection frame; represents the grid corresponding to the foreground-background segmentation result, and is 1 when the proportion of foreground is more than 50%, otherwise 0; represents the grid corresponding to the foreground-background segmentation result, and is 1 when the proportion of background is more than 50%, otherwise 0; and represent the weights of different indicators;

[0037] S34: After the network is trained, input The position of the garbage and the type of the garbage contained in the ground image collected by the monitoring device can be obtained.

[0038] Optionally, the identified data is processed in the S4 step, the type and quantity of the garbage are analyzed and uploaded to the cloud database, facilitating further data analysis and management, including:

[0039] The identification result is transmitted to the cloud database through the 5G network for recording, and the entries recorded include the position of the garbage, the type of the garbage, such as common garbage such as plastic bottles, glass, waste paper, and cigarette butts, the number of the garbage, and whether the garbage has been disposed of, etc.

[0040] Optionally, in the S5 step, according to the ground garbage information recorded in the cloud database, a route is planned to dispatch a garbage cleaning vehicle to the site to dispose of the garbage, including:

[0041] According to the unprocessed ground garbage information recorded in the cloud database, the garbage cleaning vehicle closest to the straight-line distance of the garbage is issued to clean the garbage, including:

[0042] S51: Set the current position of the garbage cleaning vehicle as the starting point, and set the position of the garbage to be cleaned as the end point, and identify the travelable section and the midway stop point of the garbage cleaning vehicle in combination with the park planning map;

[0043] ​S52: initialize the set for representing the points whose shortest path is known and their corresponding travel distance, these points are between the start point and the end point, while initializing the set for representing the points whose shortest path is not known and the travel distance from the start point to the point, the distance from the start point to the remaining stop point is added to it;

[0044] S53: find the point with the shortest path from add it to ;

[0045] S54: update the set path, calculate the sum of the distance from the newly added stop point to the stop point that has not been reached and the distance traveled by the garbage cleaning vehicle to the stop point, and compare it with the distance of the corresponding stop point that has not been reached in the current set, update it to the smaller value;

[0046] S55: repeat S53 and S54 until the garbage cleaning vehicle reaches the end point;

[0047] If the garbage cleaning vehicle assigned with the cleaning task finds other cleaning vehicles in the planned route during the travel, the cleaning task is handed over, and S51 to S55 are repeated;

[0048] After the garbage is cleaned, the result is uploaded to the cloud through the 5G network, and the processing status in the corresponding ground garbage item in the cloud database is updated.

[0049] The application also provides a 5G intelligent park ground garbage identification system, which comprises:

[0050] An image acquisition module: for acquiring ground images in the park taken by monitoring equipment;

[0051] A segmentation module: for foreground and background segmentation of the input ground image;

[0052] A garbage identification module: further identifies the position and category of potential garbage in combination with the results of the segmentation module;

[0053] A path planning module: for inputting the position of the ground garbage to be cleaned, and designing the moving route of the garbage cleaning vehicle using a path planning algorithm;

[0054] Wherein, 5G technology is used for information transmission between the image acquisition module and the segmentation module, and between the garbage identification module and the path planning module.

[0055] The application has the following advantages:

[0056] The application provides a 5G-based intelligent park ground garbage identification method and system, which uses 5G technology to transmit image data, and can ensure fast, stable and efficient data transmission. Compared with traditional technologies, the 5G technology has higher data transmission speed and larger data transmission capacity, and can effectively reduce data loss and delay in the data transmission process, and ensure data quality. The foreground-background segmentation method provided by the application provides preliminary garbage position information, which is helpful for subsequent network learning. The loss function of the network designed by the application calculates additional loss for the part of the segmentation result that is inconsistent with the actual result, which can guide the network to pay attention to the part that is easy to misclassify. The use of 5G technology can quickly upload the garbage identification result and distribute it to the garbage cleaning vehicles in the intelligent park. The garbage cleaning vehicles can automatically plan the travel route according to their starting point and the garbage position, and the transfer of cleaning tasks in the planning process can efficiently mobilize all the cleaning vehicles in the park, avoiding waste of resources. BRIEF DESCRIPTION OF DRAWINGS

[0057] Figure 1 The flowchart of an embodiment of the application. DETAILED DESCRIPTION

[0058] The application will be further described below with reference to the accompanying drawings, but the application is not limited in any way by the drawings, and any transformation or replacement based on the teaching of the application belongs to the protection scope of the application. Embodiment 1

[0059] In order to achieve the above-mentioned purpose, the application provides a 5G-based intelligent park ground garbage identification method and system, as shown in Figure 1 The method comprises the following steps:

[0060] S1: using the monitoring equipment in the park to collect ground images, so as to monitor the ground in the park in real time and facilitate fast acquisition of garbage information:

[0061] The monitoring equipment is arranged at different positions in the park to collect ground images. The collected images are transmitted to a data processing server through a 5G network for subsequent image recognition algorithm analysis.

[0062] S2: pre-processing the collected images to ensure image quality and data accuracy:

[0063] The image preprocessing includes two parts of image denoising and image foreground extraction. The image denoising is directly performed by using the median filtering method. The foreground extraction algorithm flow of the image includes:

[0064] S21: according to the maximum value and the minimum value of the image, a series of threshold values are selected to calculate the foreground and background segmentation results of multiple threshold values:

[0065] ;

[0066] ;

[0067] ;

[0068] wherein S represents a segmentation result image of image foreground and background, is a pixel in image S, b is a color channel of the image, and is one of red, green, and blue, P is a ground image, is a pixel in ground image V, (i,j) represents the horizontal and vertical coordinates of the pixel, represents the minimum value in the image, represents the maximum value in the image;

[0069] S22: Select the range of 10% of the edge of the ground image as the reference background, and compare the part coinciding with the reference background in the obtained series of foreground and background segmentation results as the similarity, and the calculation method of the similarity is:

[0070] ;

[0071] wherein B represents a reference background point coordinate set, is the number of reference background points, the function is used to identify whether the pixel in the segmentation result is background, .

[0072] S23: After obtaining the similarity of the foreground and background segmentation results under different thresholds and the reference background, the segmentation results with a similarity exceeding 50% are removed, and each pixel of the ground image is voted based on the remaining segmentation results, if the number of votes of the current pixel as foreground or background exceeds half of the total number of votes, the pixel is determined as foreground or background, and finally all segmentation images will be merged into a final foreground and background segmentation result of the ground image.

[0073] S3: Using image recognition technology, the image is classified and recognized, and the type and position of the garbage are recognized:

[0074] S31: Merge the ground image and the foreground and background segmentation result obtained in step S2 to obtain a merged image as the input image of the image recognition technology:

[0075] ;

[0076] wherein, is the foreground and background segmentation result obtained in step S2.

[0077] S32: The image is input into a Yolo-based target detection deep learning network to obtain the recognized type and position of the ground garbage, which is described by the following parameters:

[0078] ;

[0079] wherein represents the pixel position of the ground image where the center of the ground garbage detection frame is located; represents the width and height of the detection frame where the detected garbage is located; represents the probability that the garbage in the detection frame belongs to different types of garbage, and N represents the number of garbage types to be identified, represents the probability that there is no garbage in the detection frame.

[0080] S33: After the network outputs the parameters of the predicted detection frame, the loss function is calculated by comparing the actual detection frame, which is used to train the network, and the calculation method of the loss function is:

[0081] ;

[0082] ;

[0083] ;

[0084] ;

[0085] ;

[0086] wherein S represents the number of grids defined in the Yolo-based target detection deep learning network; represents 1 if there is garbage in the grid, otherwise 0; and respectively represent the center position of the detection frame where the garbage is located and the real value of its width and height, and the value predicted by the network; represents 1 if the grid contains the mth type of garbage, otherwise 0; represents 1 if there is no garbage in the grid, otherwise 0; represents the probability that the garbage in the detection frame belongs to different types of garbage, and N represents the number of garbage types to be identified, represents the probability that there is no garbage in the detection frame. represents the foreground-background segmentation result corresponding to the grid, and is 1 when the proportion of the foreground is more than 50%, otherwise 0; represents the foreground-background segmentation result corresponding to the grid, and is 1 when the proportion of the background is more than 50%, otherwise 0; , and represent different index weights, which are 3, 1 and 0.1 respectively in this embodiment.

[0087] S34: After the network completes training, input The position and type of garbage contained in the ground image collected by the monitoring device can be obtained.

[0088] S4: The identified data is processed, the type and quantity of garbage are analyzed, and uploaded to the cloud database for further data analysis and management:

[0089] The recognition result is transmitted to the cloud database through the 5G network for recording. The recorded entries include the position of the garbage, the type of the garbage, such as plastic bottles, glass, waste paper, and cigarette butts, the number of the garbage, and whether the garbage has been disposed of, etc.

[0090] S5: According to the ground garbage information recorded in the cloud database, the route is planned to dispatch the garbage cleaning car to the site to dispose of the garbage:

[0091] S51: Set the current garbage cleaning car position as the starting point, and the garbage position to be cleaned as the end point, and identify the garbage cleaning car travelable route and the midway stop points combined with the park planning map;

[0092] S52: Initialize the set to represent the points of the known shortest path and their corresponding travel distances, which are located between the starting point and the end point, and initialize the set to represent the points whose shortest path is unknown and the travel distance from the starting point to the point, and add the distance from the starting point to the remaining stop points to it;

[0093] S53: Find the point with the shortest path from and add it to ;

[0094] S54: Update the path of the set , calculate the sum of the distance from the newly added stop point to the stop points that have not been reached and the distance traveled by the garbage cleaning car to the stop point, and compare it with the corresponding distance of the stop points that have not been reached in the current set, and update it to the smaller value;

[0095] S55: Repeat S53 and S54 until the garbage cleaning car reaches the end point;

[0096] If the garbage cleaning car assigned with the cleaning task finds other cleaning cars in the planned route during the travel process, the cleaning task is handed over, and S51 to S55 are repeated;

[0097] After the garbage is cleaned, the result is uploaded to the cloud through the 5G network, and the disposal status of the corresponding ground garbage entry in the cloud database is updated. Embodiment 2

[0098] The application also discloses a 5G-based intelligent park ground garbage identification system, which comprises:

[0099] An image acquisition module is configured to acquire ground images in the park captured by the monitoring device.

[0100] A segmentation module is configured to perform foreground and background segmentation on the input ground images.

[0101] A garbage identification module is configured to further identify the position and category of potential garbage in combination with the result of the segmentation module.

[0102] A path planning module is configured to take the position of the ground garbage to be cleaned as input and design a moving route of the garbage cleaning vehicle by using a path planning algorithm.

[0103] The image acquisition module and the segmentation module, and the garbage identification module and the path planning module are connected by using 5G technology for information transmission.

[0104] It should be noted that the above-mentioned application embodiment serial numbers are only for description, and do not represent the advantages and disadvantages of the embodiments. Moreover, the terms "include", "contain" or any other variants thereof in this paper are intended to cover non-exclusive inclusion, so that the process, device, article or method including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or includes elements inherent to such process, device, article or method. Without more limitations, the element defined by the statement "including a" does not exclude the presence of another identical element in the process, device, article or method including the element.

[0105] From the above description of the embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment method can be realized by software and necessary general hardware platform, of course, it can also be realized by hardware, but in many cases, the former is a better embodiment. Based on such understanding, the technical solutions of the application can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes a plurality of instructions for making a terminal device (which can be a mobile phone, computer, server, or network device, etc.) execute the method described in each embodiment of the application.

[0106] The above is only the preferred embodiment of the application, and does not limit the patent scope of the application, and any equivalent structure or equivalent flow transformation made by using the content of the application specification and drawings, or direct or indirect application in other related technical fields, are also included in the patent protection scope of the application.

Claims

1. A method for identifying ground garbage in a 5G smart park, characterized in that: It comprises the following steps: S1: Collecting ground images using monitoring devices in the park to monitor the ground in the park in real time, facilitating quick access to garbage information; S2: Preprocessing the collected images to ensure image quality and data accuracy; The preprocessing of the collected images in the S2 step to ensure image quality and data accuracy comprises: The preprocessing of the images comprises two parts of image denoising and image foreground extraction, the image denoising is directly performed using the median filtering method, and the image foreground extraction algorithm process comprises: S21: Selecting a series of threshold values according to the maximum and minimum values of the image to calculate the foreground and background segmentation results of the multiple threshold values: t e [min(V b )+1,max(V b )-1] ; k = t - min(V b ); where S represents a segmentation result image of the image foreground and background, τ is a pixel in the image S, b is a color channel of the image, which is one of red, green and blue, P is a ground image, υ is a pixel in the ground image V, (i,j) represents the horizontal and vertical coordinates where the pixel is located, min(V b ) represents the minimum value in the image, max(V b ) represents the maximum value in the image, t is a threshold value for distinguishing the foreground and background in the ground image; k is a relative offset of the current threshold value t from the minimum pixel value. S22: Selecting the range of 10% of the edge of the ground image as the reference background, comparing the part coinciding with the reference background in the obtained series of foreground and background segmentation results as the similarity, and the calculation method of the similarity is: Wherein, B represents the reference background point coordinate set, sum(B) is the reference background point number, The function is used to identify whether the pixel in the segmentation result is background or not, (i, j) ∈ B; S23: After obtaining the similarity of the foreground and background segmentation results under different threshold values and the reference background, the segmentation results with a similarity exceeding 50% are removed, the remaining segmentation results are used to vote for each pixel of the ground image, if the number of votes for the current pixel as foreground or background exceeds half of the total number of votes, the pixel is determined as foreground or background, and finally all segmentation images will be merged into a final foreground and background segmentation result of the ground image; S3: Using image recognition technology to classify and recognize the images, and identifying the types and positions of the garbage; S4: Processing the recognized data, analyzing the types and quantities of the garbage, and uploading to the cloud database for further data analysis and management; S5: According to the ground garbage information recorded in the cloud database, planning the route and dispatching the garbage cleaning vehicle to process the garbage. 2.The 5G-based smart park ground litter identification method of claim 1, wherein, The image recognition technology used in the S3 step to classify and recognize the images and identify the positions and types of the garbage comprises: S31: Merging the ground image and the foreground and background segmentation results obtained in the step S2 to obtain a merged image as an input image of the image recognition technology: wherein, is the foreground-background segmentation result obtained in step S2; S32: Inputting the image into a target detection deep learning network based on Yolo to obtain the recognized types and positions of the ground garbage, which are described by the following parameters: (c x c y ,c w ,c h ,p1,p2,…p N ,p notrash ); wherein c x , c y represents the pixel position of the ground image where the center of the ground garbage detection frame is located; c w , c h represents the width and height of the detection frame where the detected garbage is located; p1, p2,... p N represents the probability that the garbage in the detection frame belongs to different categories of garbage, N represents the number of garbage categories to be identified, p notrash represents the probability that there is no garbage in the detection frame; S33: Comparing the parameters contained in the predicted detection frame output by the network with the actual detection frame to calculate a loss function for training the network, and the calculation method of the loss function is: L = γ1L1 + γ2L2 + γ3L3; S represents the number of grids defined in the Yolo-based target detection deep learning network; represents 1 if there is garbage in the grid, otherwise 0; c l,x l,y l,w l,h l，x l，y l，w l，h respectively represent the center position of the detection frame where the garbage is located and the real value of the width and height of the detection frame when there is garbage in the grid, and the network predicted value; represents 1 if the grid contains the mth type of garbage, otherwise 0; represents 1 if there is no garbage in the grid, otherwise 0; p1, p2, … p N represents the probability that the garbage in the detection frame belongs to different types of garbage, N represents the number of garbage types to be identified, and p notrash represents the probability that there is no garbage in the detection frame; represents the grid corresponding to the foreground-background segmentation result, and is 1 when the proportion of foreground is more than 50%, otherwise 0; represents the grid corresponding to the foreground-background segmentation result, and is 1 when the proportion of background is more than 50%, otherwise 0; γ1, γ2 and γ3 represent different index weights, c l,e e∈{x,y,w,h}, so in the operation c l,e will first become c l,x , c l,y , c l,w , c l,h , and then participate in the operation, and c l,x , c l,y , c l,w , c l,h represents the network predicted value of the center position of the detection frame where the garbage is located and the width and height of the detection frame when there is garbage in the grid, is the p notrash power of a natural constant;​​​​​​​ S34: After the network is trained, inputting I can obtain the position and type of the garbage contained in the ground image collected by the monitoring device. 3.The 5G-based smart park ground litter identification method of claim 1, wherein, The processing of the recognized data in the S4 step, the analysis of the types and quantities of the garbage, and the uploading to the cloud database for further data analysis and management comprise: The recognition results are transmitted to the cloud database through the 5G network for recording, and the recorded entries include the position of the garbage, the type of the garbage, the number of the garbage, and whether the garbage has been processed. 4.The 5G-based smart park ground litter identification method of claim 1, wherein, The S5 step of planning the route and dispatching the garbage cleaning vehicle to process the garbage according to the ground garbage information recorded in the cloud database comprises: According to the unprocessed ground garbage information recorded by the cloud database, the garbage cleaning vehicle closest to the garbage in the straight-line distance is issued to clean the garbage, including: S51: set the current garbage cleaning vehicle position as the starting point, and the garbage to be cleaned position as the end point, identify the garbage cleaning vehicle travelable route and the midway stopover points in combination with the park planning map; S52: initialize the set φ to represent the points of the known shortest path and the corresponding travel distance, which are located between the starting point and the end point, and initialize the set ψ to represent the points of the unknown shortest path and the travel distance from the starting point to the point, add the distance from the starting point to the remaining stopover points to the set; S53: find the point with the shortest path from ψ and add it to φ; S54: update the ψ set path, calculate the sum of the distance from the newly added stopover point to the stopover points that have not been reached and the distance traveled by the garbage cleaning vehicle to the stopover point, and compare it with the corresponding distance of the stopover points that have not been reached in the current ψ set, and update it to a smaller value; S55: repeat S53 and S54 until the garbage cleaning vehicle reaches the end point; If the garbage cleaning vehicle assigned with the cleaning task finds other cleaning vehicles in the planning route during the travel process, the cleaning task is handed over, and S51 to S55 are repeated; After the garbage is cleaned, the result is uploaded to the cloud through the 5G network to update the processing status of the corresponding ground garbage entry in the cloud database.

5. A 5G smart park ground garbage identification system based on, Including: Image acquisition module: used for acquiring the ground image in the park shot by the monitoring device; Segmentation module: used for foreground and background segmentation of the input ground image; Garbage recognition module: further recognize the position and category of potential garbage in combination with the result of the segmentation module; Path planning module: used for designing the moving route of the garbage cleaning vehicle with the position of the ground garbage to be cleaned as the input, using path planning algorithm; Wherein, 5G technology is used for information transmission between the image acquisition module and the segmentation module, and between the garbage recognition module and the path planning module.

Images