A method and apparatus for cleaning and fusing roadside parking trajectory data

Abstract

This invention discloses a method and apparatus for cleaning and fusing roadside parking trajectory data, relating to the technical field of parking management. The method includes: acquiring current inspection data, obtained by a patrol vehicle capturing images of all parking spaces within a parking area, including license plate recognition data, parking space status (parking space number, occupied / vacant), and additional information (detour signs, filtered license plates, and a list of passing vehicles); performing data cleaning: determining whether a parking space is occupied; if occupied, comparing the license plate recognition data with the list of vehicles present; if no matching license plate is found and no detour is taken, the data is accepted; if no matching license plate is found but a detour sign is present, the data is cleaned by combining multiple pieces of information; if a vehicle with the same license plate is present, the data is cleaned after determining adjacent parking spaces. If the parking space is not occupied, the data is cleaned based on the current inspection data and the filtered license plate list. Finally, parking trajectory fusion is performed based on the cleaned data. This invention can clean inspection data in a patrol system, accurately reconstructing the actual parking trajectory of vehicles.

Description

Technical Field

[0001] This invention belongs to the technical field of parking management, and specifically relates to a method and apparatus for cleaning and fusing roadside parking trajectory data. Background Technology

[0002] In modern cities, using road patrol vehicles for parking management is an important part of traffic control. Common problems with roadside parking include improper parking, vehicles obstructing traffic, and tall buildings or trees along the roadside, which cause unstable GPS signals, resulting in blind spots and inconsistencies in the patrol vehicle's inspection, thus making it impossible to accurately reconstruct the parking trajectory.

[0003] Specifically, existing technologies used in the on-street parking industry have the following problems: (1) Instability of GPS signals leads to errors in the judgment of parking space information by inspection vehicles. In urban environments, tall buildings and trees commonly found on both sides of the road can hinder the stability of GPS signals, causing inspection vehicles to be unable to accurately locate when collecting parking information, resulting in blind spots or discontinuous tracks. This directly affects the inspection vehicle's accurate monitoring of parking spaces and makes it impossible to accurately reconstruct the parking trajectory of vehicles in the parking spaces; (2) Obstruction and improper parking issues. Vehicles parked on the roadside may obstruct other vehicles or parking spaces, especially in cases of improper parking. Existing inspection vehicles often have difficulty handling obstruction issues when collecting information, leading to a decrease in the accuracy of parking information and affecting the subsequent inspection analysis and management effects; (3) Lack of intelligent data cleaning mechanisms. In existing technologies, the erroneous information collected by inspection vehicles cannot be effectively cleaned and corrected automatically. The data collected by inspection vehicles due to environmental factors, obstructions, etc., contains a large amount of noise, and the lack of automated data cleaning mechanisms results in data quality that cannot meet the needs of refined parking management.

[0004] The causes of these problems are as follows: ① The road is located between tall buildings or is obstructed by trees or other objects on both sides of the road; ② The limitations of GPS technology itself; ③ The problems of license plate obstruction and improper parking; ④ The limitations of the data collection method of patrol vehicles; ⑤ The lack of data fusion technology; ⑥ The noise problem in data collection; ⑦ The lack of an automated data cleaning and error correction system.

[0005] The above reasons reveal the multiple challenges faced by existing road patrol vehicles in urban parking management. These mainly include technological limitations (such as weak GPS signals), insufficient data processing capabilities (such as a lack of multi-source fusion), the complexity of the parking environment (such as occlusion issues), and inadequate data cleaning and intelligent processing capabilities. These factors combined make it difficult for existing patrol systems to accurately reconstruct the actual parking trajectories of vehicles. Summary of the Invention

[0006] To address the aforementioned issues, this invention proposes a method and apparatus for cleaning and fusing roadside parking trajectory data, which can clean the inspection data in the inspection system and accurately reconstruct the actual parking trajectory of the vehicle.

[0007] In a first aspect, the present invention provides a method for cleaning and fusing roadside parking trajectory data, applied to a server, the method comprising:

[0008] Acquire current inspection data; wherein, the current inspection data is obtained by the inspection vehicle capturing images of the target parking space, and the inspection vehicle captures images of all parking spaces in the parking area once within the current capture cycle; the current inspection data includes license plate recognition data, parking space status, and additional information; the additional information includes detour signs, filtered license plates, and a list of passing vehicles; the parking space status includes parking space number, occupied, and vacant;

[0009] Data cleaning is performed based on the current inspection data to obtain cleaned data, including:

[0010] Determine whether the target berth is occupied based on the berth status;

[0011] If the target parking space is occupied, it is determined whether there is a vehicle with the same license plate based on the license plate recognition data and the list of vehicles in the parking area; wherein, the list of vehicles in the parking area is stored in the server and is a record table of vehicles entering the parking area during the current capture period;

[0012] If there is no vehicle with the same license plate on site and it does not have a detour sign, the current inspection data is accepted.

[0013] If there is no vehicle with the same license plate on site and it carries the detour sign, then data cleaning is performed by combining whether there is a vehicle on site at the target parking space, the additional information, and the license plate recognition data.

[0014] If there are vehicles with the same license plate on site, then adjacent parking spaces are determined, and data cleaning is performed based on the results of the adjacent parking space determination.

[0015] If the target parking space is not occupied, data cleaning is performed based on the current inspection data and the filtered license plate list;

[0016] The parking trajectory is fused based on the cleaned data to obtain a parking trajectory that can reconstruct the actual parking situation.

[0017] In an optional implementation, the current inspection data further includes captured GPS data, GPS signal strength, inspection vehicle riding type, and capture time; wherein, the license plate recognition data includes the target vehicle's license plate number, license plate color, and recognition reliability, and the inspection vehicle riding type includes normal and detour.

[0018] In an optional implementation, the license plate recognition data further includes a license plate image; the data cleaning process, which combines the presence of a vehicle at the target parking space, the additional information, and the license plate recognition data, includes:

[0019] Determine whether there is a vehicle at the target berth based on the berth number and the list of vehicles on site;

[0020] If there is a vehicle in the target parking space, data cleaning is performed based on the list of passing vehicles and the license plate image;

[0021] If there is no vehicle at the target berth, the current inspection data is accepted.

[0022] In an optional implementation, the data cleaning based on the list of passing vehicles and the license plate images includes:

[0023] Determine whether the list of passing vehicles contains the license plate information of the vehicle at the target parking space;

[0024] If the license plate information of the vehicle in the target parking space is included, the license plate recognition data is modified according to the license plate information of the vehicle in the target parking space.

[0025] If the license plate information of the vehicle in the target parking space is not included, then determine whether the number of pixels in the license plate image is greater than a preset pixel threshold.

[0026] If the value exceeds a preset pixel threshold, the current inspection data is discarded.

[0027] If the value is not greater than a preset pixel threshold, the current inspection data is accepted.

[0028] In an optional implementation, the step of determining adjacent berths and cleaning the data based on the adjacent berth determination results includes:

[0029] Determine whether the berth number of the target berth stored in the server and the berth number of the target berth captured in this snapshot are adjacent berth numbers;

[0030] If the parking space numbers are adjacent, the on-site confirmation record of the target vehicle is queried, and the data is cleaned based on the query results.

[0031] If it is not an adjacent berth number, then the current inspection data is accepted.

[0032] In an optional implementation, the step of querying the presence confirmation record of the target vehicle and cleaning the data based on the query results includes:

[0033] Determine whether the target vehicle has a record of being present based on its license plate number and the list of vehicles present;

[0034] If there is a record of presence, the parking space number in the parking space status will be modified according to the parking space number of the target vehicle. Also, if there is uncleaned data in the current capture cycle and the corresponding data of the target vehicle is still stored in the parking vehicle list, the corresponding data of the target vehicle stored in the parking vehicle list will be invalidated.

[0035] If there is no record of presence, the current inspection data is accepted.

[0036] In an optional implementation, the additional information further includes berth type, which includes parallel berths, perpendicular berths, and angled berths; if the target berth is not occupied, data cleaning is performed based on the current inspection data and the filtered license plate list, including:

[0037] Determine if the filtered license plate information for the target vehicle is present in the filtered license plate list;

[0038] If there is license plate information corresponding to the target filtering vehicle, then it is determined whether the target filtering vehicle is an on-site vehicle and has an on-site confirmation mark based on the license plate information. The on-site confirmation mark is obtained from the on-site vehicle record and is used to indicate that the license plate of the target filtering vehicle has been captured multiple times at the target parking space.

[0039] If the vehicle is present and has an presence confirmation mark, then determine whether the parking space number corresponding to the filtered license plate information is consistent with the parking space number in the parking space status.

[0040] If they match, the license plate recognition data is modified according to the filtered license plate information, and the parking space status is changed to occupied.

[0041] If there is a discrepancy, the current inspection data is accepted.

[0042] If the vehicle is not present, then the current inspection data is accepted.

[0043] If there is no license plate information corresponding to the target vehicle, then determine the parking space type;

[0044] If it is a parallel berth and there is no detour sign, or if it is a perpendicular berth and there is no vehicle on site, or if it is an angled berth and there is no vehicle on site, then the current inspection data is accepted.

[0045] If it is a parallel parking space with the detour sign, or if it is a perpendicular parking space with a vehicle in the parking space, or if it is an angled parking space with a vehicle in the parking space, then determine whether the target vehicle's license plate number is in the list of passing vehicles.

[0046] If the license plate number of the target vehicle is available, the current inspection data is modified based on the license plate recognition data of the target vehicle.

[0047] If the target vehicle's license plate number is not available, the current inspection data will be invalidated.

[0048] In an optional implementation, the step of fusing parking trajectories based on the cleaned data to obtain a parking trajectory capable of reconstructing the parking facts includes:

[0049] Based on the berth status in the cleaned data, determine whether it is occupied;

[0050] If the parking space is occupied, check the list of vehicles in the parking area to determine if there is a parking record.

[0051] If there is a parking record, then determine the license plate number and target parking space in the cleaned data.

[0052] Check if the license plate numbers of the vehicles parked on the vehicle are consistent;

[0053] If they match, confirm the presence of the vehicle parked at the target parking space; proceed to output.

[0054] The steps involved in parking trajectory tracking;

[0055] If there is a discrepancy, the parking space at the target parking spot will be removed, and the parking space will be re-entered based on the data after cleaning.

[0056] If there is no parking record, then entry is performed based on the cleaned data; proceed to the step of outputting the parking trajectory;

[0057] If the parking space is unoccupied, then determine whether the target parking space is occupied.

[0058] If there is a parking space, the parking space in the target parking space will be exited; proceed to the step of outputting the parking trajectory.

[0059] If the vehicle is not parked, proceed to the step of outputting the parking trajectory;

[0060] Output parking trajectory.

[0061] In an optional implementation, the step of determining whether there is a vehicle with the same license plate based on the license plate recognition data and the list of vehicles present further includes:

[0062] If the current inspection data already exists, it is determined based on historical inspection data. If the current inspection data already exists, the berth status in the current inspection data is modified to an empty berth. The historical inspection data is stored in the server and is the inspection data captured by the inspection vehicle in at least one historical capture cycle.

[0063] Secondly, the present invention provides a data cleaning and fusion device for roadside parking trajectories, comprising:

[0064] The data acquisition module is used to acquire current inspection data; wherein, the current inspection data is obtained by the inspection vehicle capturing images of the target parking space, and the inspection vehicle captures images of all parking spaces in the parking area once within the current capture cycle; the current inspection data includes license plate recognition data, parking space status, and additional information; the additional information includes detour signs, filtered license plates, and a list of passing vehicles; the parking space status includes parking space number, occupied, and vacant;

[0065] The judgment and cleaning module is used to clean the data based on the current inspection data to obtain cleaned data;

[0066] The trajectory fusion module is used to fuse parking trajectories based on the cleaned data to obtain parking trajectories that can reconstruct the actual parking situation.

[0067] The judgment and cleaning module includes:

[0068] The first berth occupancy determination module is used to determine whether the target berth is occupied based on the berth status.

[0069] The first parking space occupancy module is used to determine whether there is a vehicle with the same license plate if the target parking space is occupied, based on the license plate recognition data and the list of vehicles in the parking area; if there is no vehicle with the same license plate and no detour sign, the current inspection data is accepted; if there is no vehicle with the same license plate and the vehicle has the detour sign, data cleaning is performed by combining whether there is a vehicle in the target parking space, the additional information, and the license plate recognition data; if there is a vehicle with the same license plate, adjacent parking spaces are judged, and data cleaning is performed based on the adjacent parking space judgment result; wherein, the list of vehicles in the parking area is stored in the server, and the list of vehicles in the parking area is an entry record table of vehicles in the parking area during the current capture period;

[0070] The first unoccupied berth module is used to perform data cleaning based on the current inspection data and the filtered license plate list if the target berth is not occupied.

[0071] Thirdly, the present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method described in any of the foregoing embodiments.

[0072] Fourthly, the present invention provides a computer-readable medium having processor-executable non-volatile program code, the program code causing the processor to perform the method described in any of the foregoing embodiments.

[0073] The beneficial effects of the technical solution provided by the embodiments of the present invention are as follows: The data cleaning and fusion method and apparatus for roadside parking trajectory of the present invention combines the license plate recognition data, parking space status and additional information in the inspection data with the historical inspection data stored in the server to make a comprehensive judgment, thereby correcting the inspection data in the current capture cycle and improving the accuracy of the parking trajectory. Attached Figure Description

[0074] Figure 1 A flowchart illustrating a method for cleaning and fusing roadside parking trajectory data provided in an embodiment of the present invention;

[0075] Figure 2 This is another flowchart illustrating a method for cleaning and fusing roadside parking trajectory data provided in an embodiment of the present invention; it aims to demonstrate the cleaning steps for parking space occupancy.

[0076] Figure 3 This is another flowchart illustrating a method for cleaning and fusing roadside parking trajectory data provided in an embodiment of the present invention; it aims to demonstrate the cleaning steps for unoccupied parking spaces.

[0077] Figure 4 This is another flowchart illustrating a method for cleaning and fusing roadside parking trajectory data provided in an embodiment of the present invention; it aims to demonstrate the steps of parking trajectory fusion.

[0078] Figure 5 A system schematic diagram of a data cleaning and fusion device for roadside parking trajectories provided in an embodiment of the present invention;

[0079] Figure 6 A schematic diagram of the system principle of an electronic device provided in an embodiment of the present invention.

[0080] In the diagram: 1000 - Data acquisition module; 2000 - Judgment and cleaning module; 2100 - First berth occupancy judgment module; 2200 - First berth occupancy module; 2220 - First on-site vehicle judgment module; 2230 - No on-site vehicle detour module; 2240 - No on-site vehicle detour module; 2250 - On-site vehicle module; 2300 - First berth unoccupied module; 3000 - Trajectory fusion module; 400 - Electronic equipment; 401 - Communication interface; 402 - Processor; 403 - Memory; 404 - Bus. Detailed Implementation

[0081] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0082] See Figures 1-4 This embodiment provides a method for cleaning and fusing roadside parking trajectory data, see [link to relevant documentation]. Figure 1The method in this embodiment includes the following steps S1000 to S3000.

[0083] Step S1000: Obtain current inspection data; wherein, the current inspection data is obtained by the inspection vehicle capturing images of the target parking space, and the inspection vehicle captures images of all parking spaces in the parking area once within a preset period; the current inspection data includes license plate recognition data, parking space status and additional information; the additional information includes detour signs, filtered license plates and a list of passing vehicles; the parking space status includes parking space number, occupied and vacant.

[0084] Specifically, the inspection vehicle periodically captures images of all vehicles and parking spaces within the parking area, and uploads the images to the server. Steps S1000 to S3000 in this embodiment clean the data from a single image captured at one parking space. For the remaining parking spaces, steps S1000 to S3000 are repeated. This embodiment uses one parking space as the target parking space for explanation. The server here is a system application deployed to process inspection data, such as a parking management system. Generally, the inspection vehicle uploads inspection data via wireless networks, such as mobile networks (e.g., 4G or 5G communication modules) and short-range wireless networks (e.g., Wi-Fi).

[0085] It should also be noted that, such as Figure 4 As shown, before executing the method of this embodiment, the method further includes the following steps Z100 to Z300, wherein steps Z100 to Z200 are data preparation steps, and step Z300 is a data acquisition step.

[0086] Step Z100: Calibrate all berth GPS locations.

[0087] Before deploying the system application on the server, staff need to accurately calibrate the GPS coordinates of all parking spaces within the operational parking area to ensure that the vehicle parking trajectory accurately matches the actual parking space location. The deployed GPS coordinates are then saved to the server.

[0088] Step Z200: Configure the vehicle terminal system (this system requires 2 to 4 cameras).

[0089] Here, the vehicle-mounted terminal collects vehicle information, parking space status information, and GPS data in real time through integrated hardware modules, and uploads them to the cloud for processing and storage. These hardware modules include at least two to four high-definition cameras, a GPS acquisition module, and a 4G communication module.

[0090] Step Z300: The inspection vehicle periodically collects inspection data.

[0091] Specifically, the inspection data collected by the patrol vehicle is multi-source data. The data captured by the patrol vehicle is periodically uploaded to the server, and the server receives data from various data sources output by the patrol vehicle. Typically, the patrol vehicle's capture cycle is 30 minutes, meaning one cycle is 30 minutes, and each berth's information is captured once every 30 minutes. This data provides a complete information foundation for subsequent data processing and analysis on the server side.

[0092] The inspection data captured by the inspection vehicle and uploaded to the server (i.e., the key information captured by the inspection vehicle each time) includes the following items ① to ⑥.

[0093] ① License plate recognition data, including at least the license plate number, license plate color, and recognition reliability of the target vehicle (e.g., vehicle A);

[0094] ② Capture GPS data and GPS signal strength;

[0095] ③ Berth status, including at least berth number, occupied and available;

[0096] ④ The patrol vehicle riding types should include at least normal and detour routes;

[0097] ⑤ Capture time;

[0098] ⑥ Attachment information, including at least detour signs, filtered license plates, and a list of passing vehicles. Among them, filtered license plates are vehicles identified and determined to be outside the designated lanes during the inspection process; passing vehicle license plates are vehicles identified and determined to be passing vehicles (short-term stops) during the inspection process.

[0099] Preferably, before step S1000, the method further includes step Z400, data preprocessing.

[0100] Data preprocessing includes data synchronization and noise reduction. During preprocessing, the timestamps of the inspection data must be consistent with the server. Timestamps from different data sources can be matched within a range. If the timestamps differ significantly, the server and the inspection vehicle terminal are notified to synchronize their times to ensure consistency between the data sources and the server. Data preprocessing ensures that the inspection data uploaded to the server is accurate and synchronized, and removes potential noise and inconsistencies.

[0101] Step S2000: Perform data cleaning based on the current inspection data to obtain cleaned data.

[0102] Further, see Figure 1 Step S2000 includes steps S2100 to S2300.

[0103] Step S2100: Determine whether the target berth is occupied based on the berth status. That is, if the berth information in the current inspection data indicates that the berth is occupied, proceed to step S2200; if the berth information in the current inspection data indicates that the berth is not occupied, proceed to step S2300.

[0104] Step S2200, see Figure 1 and Figure 2 For vehicle A with license plates, if the target parking space is occupied, steps S2210 to S2260 are executed. These steps clean and correct the data captured by the inspection vehicle to improve the accuracy of identifying parking space status and parking behavior.

[0105] Step S2210: Determine whether the current inspection data already exists based on historical inspection data. Historical inspection data is stored on a server and consists of inspection data captured by the inspection vehicle in at least one historical capture cycle. This historical inspection data can be the inspection data captured in the most recent three rounds (three cycles). In this embodiment, the first cleaning and correction is performed using the historical inspection data captured in the most recent three rounds. It is determined whether the filtered license plate list in this historical data contains the license plate number of target vehicle A. If it does, it indicates that the current inspection data already exists; otherwise, it indicates that the current inspection data does not exist.

[0106] Step S2211: If the current inspection data already exists, then the capture is determined to be a false capture. The berth status in the current inspection data is then changed to an empty berth, and the process proceeds to step S2260.

[0107] If the previous inspection data is not available, proceed to the next cleaning and correction step and execute step S2220.

[0108] Step S2220: Determine whether there is a vehicle with the same license plate based on the license plate recognition data and the list of vehicles on site. Obtain the license plate number of vehicle A from the license plate recognition data in the current inspection data, and determine whether the license plate number of the target vehicle is in the list of vehicles on site; if it is, it indicates that there is a vehicle with the same license plate on site; if it is not, it indicates that there is no vehicle with the same license plate on site.

[0109] In step S2230, if there is no vehicle with the same license plate on site and no detour sign is displayed, the current inspection data is accepted; proceed to step S2260 to end the process. The absence of a detour sign indicates that the inspection vehicle and the parking space are within an acceptable distance range, and the accuracy of the captured data within this distance range meets the requirements.

[0110] In step S2240, if there is no vehicle with the same license plate and the vehicle has a detour sign, then data cleaning is performed by combining whether there is a vehicle in the target parking space, additional information, and license plate recognition data.

[0111] Specifically, step S2240 includes the following steps S2241 to S2243.

[0112] Step S2241: Determine whether a vehicle exists at the target berth based on the berth number and the list of vehicles on site. Specifically, this is done by checking the berth number in the current inspection data. If the berth number is in the list of vehicles on site, it indicates that the license plate captured in this round is present in the server's records; otherwise, it indicates that the license plate is not present in the server's records. The list of vehicles on site refers to the data received by the server from the inspection vehicles, determining whether they have entered the berth. If they have entered, an entry record is generated and stored in the list.

[0113] Step S2242: If there are vehicles in the target parking space, data cleaning is performed based on the list of passing vehicles and license plate images. Specifically, step S2242 includes steps S2242-1 to S2242-3.

[0114] Step S2242-1: Determine whether the list of passing vehicles contains the license plate information of the vehicle at the target parking space. That is, obtain the license plate information of the vehicle at the target parking space and determine whether the list of passing vehicles contains the license plate number of target vehicle A.

[0115] Step S2242-2: If the license plate information of the vehicle in the target parking space is included, then modify the license plate recognition data according to the license plate information of the vehicle in the target parking space.

[0116] Step S2242-3: If the license plate information of the vehicle in the target parking space is not included, determine whether the number of pixels in the license plate image is greater than a preset pixel threshold. The preset pixel threshold is 400–700 dips, preferably 500 dips.

[0117] Step S2242-3-1: If the value is greater than the preset pixel threshold, the current capture is determined to be an off-line multi-detection, and the current inspection data is invalidated; proceed to step S2260 to end the process.

[0118] In step S2242-3-2, if the data is not greater than the preset pixel threshold, the current inspection data is accepted, and the process proceeds to step S2260 to end the process.

[0119] In step S2243, if there is no vehicle in the target parking space, the current inspection data is accepted, and the process proceeds to step S2260 to end the process.

[0120] In step S2250, if there are vehicles with the same license plate present, adjacent parking spaces are determined, and data cleaning is performed based on the results. The data collected by the patrol vehicle is not always accurate; for example, the patrol vehicle may mistakenly identify a vehicle as being associated with an adjacent parking space. To compensate for this inaccuracy, corrections are made in steps S2251 to S2253.

[0121] Step S2251: Determine whether the berth number of the target berth stored in the server and the berth number of the target berth captured in this snapshot are adjacent berth numbers; if the last digits of the two berth numbers are adjacent and the difference is no more than 1, they are considered to be adjacent berth numbers. For example, if the stored berth number is 001 and the berth number captured in this snapshot is 002, then they are considered to be adjacent berth numbers.

[0122] Step S2252: If the parking space number is adjacent, perform an on-site confirmation record query for the target vehicle and clean the data based on the query results. Further, step S2252 includes steps S2252-1 to S2252-3.

[0123] Step S2252-1: Determine whether the target vehicle has an presence record based on its license plate number and the list of vehicles present, i.e., determine whether target vehicle A has an presence confirmation record.

[0124] Step S2252-2: If there is an on-site record, modify the parking space number in the parking space status according to the parking space number of the target vehicle. Also, if there is uncleaned data within the current capture cycle, and the corresponding data for the target vehicle is still stored in the on-site vehicle list, invalidate the corresponding data for the target vehicle stored in the on-site vehicle list, and proceed to step S2260 to end the process. Essentially, modify the parking space number of the current capture data to match the parking space number in the current on-site record of that license plate. After modification, check if there are other capture data with the same parking space number in the on-site record of that license plate in the uncleaned capture data of this round (within 2 minutes of the current server time). If so, invalidate them.

[0125] In step S2252-3, if there is no on-site record, the current inspection data is accepted, and the process is transferred to step S2260 to end the process.

[0126] Step S2253: If it is not an adjacent berth number, then the current inspection data is accepted.

[0127] Step S2260: End the cleaning process.

[0128] In summary, see Figure 2The data cleaning steps for parking space occupancy are as follows: First, determine if the target vehicle A was included in any of the filtered license plates received in the previous three rounds for that parking space. If so, modify the current inspection data to an empty parking space. If not, determine if there is a vehicle with the same license plate present. If so, determine if it is an adjacent parking space. If there is no adjacent parking space, accept and output the current inspection data. If it is an adjacent parking space, determine if there is an presence confirmation record for that license plate in the current presence record. If so, modify the parking space number in the current inspection data to match the parking space number in the current presence record of that license plate. After modification, check if there are other inspection data for the parking space number in the presence record of that license plate in the current round of uncleaned capture data. If so, discard it.

[0129] If no vehicle with the same license plate is present, determine whether to detour. If not, accept and output the current inspection data. If detour, check if a vehicle exists at the current inspection data parking space number. If it exists, compare the license plate with the passing vehicle list in the supplementary information. If it exists, modify the license plate in the current inspection data to match the license plate of the vehicle present and output the result. If it does not exist, check the license plate image pixels. If the pixels are greater than 500, determine that this capture is an off-line over-detection and invalidate the current inspection data. If the pixels are less than or equal to 500, accept and output the current inspection data.

[0130] Based on the above analysis and judgment, the corrected data of the parking space occupancy capture is finally output for the subsequent step S3000 to perform parking trajectory fusion.

[0131] Step S2300: If the target parking space is not occupied, data cleaning is performed based on the current inspection data and the filtered license plate list. This step specifically includes steps S2310 to S2340, which clean and correct the data captured by the inspection vehicle to improve the accuracy of identifying parking space status and parking behavior.

[0132] Step S2310: Determine if there is any filtered license plate information corresponding to the target filtered vehicle in the filtered license plate list. When the server receives the captured data, it determines the filtered license plate information in the additional information of the current inspection data obtained in this capture. If filtered license plate B exists, it means that this capture may have misjudged a vehicle occupying a parking space as an empty space, and then proceeds to step S2320.

[0133] Step S2320: If there is license plate information corresponding to the target filtering vehicle, then determine whether the target filtering vehicle is an on-site vehicle with an on-site confirmation identifier based on the license plate information. The on-site confirmation identifier is obtained from the on-site vehicle record and indicates that the license plate of the target filtering vehicle has been captured multiple times at the target parking space. This embodiment uses target filtering vehicle B as an example. If there is a license plate B, and the system shows that the vehicle is on-site and has undergone on-site confirmation (the license plate has been captured multiple times at this parking space in the current on-site vehicle record), and the current on-site record parking space number matches the current captured parking space number, then the system determines that the capture is incorrect and corrects the empty parking space capture data to capture license plate B, changing the parking space status to occupied. If the current on-site record parking space number does not match the current captured parking space number, then the captured data is accepted, and the cleaned data is finally output, ending the process. The specific principle is shown in steps S2321-S2322.

[0134] Step S2321: If the vehicle is present and has a presence confirmation mark, determine whether the parking space number corresponding to the filtered license plate information is consistent with the parking space number in the parking space status.

[0135] Step S2321-1: If they match, modify the license plate recognition data according to the filtered license plate information and change the parking space status to occupied; proceed to step S2340 to end the process.

[0136] If there is a discrepancy in step S2321-2, then the current inspection data is accepted; proceed to step S2340 to end the process.

[0137] In step S2322, if the vehicle is not present, the current inspection data is accepted; proceed to step S2340 to end the process.

[0138] In step S2330, if there is no filtered license plate information corresponding to the target filtered vehicle, the parking space type is determined. If there is no filtered license plate B, but the system does not have a vehicle present for that license plate, or there is a vehicle present but its presence has not been confirmed (the license plate has been captured multiple times in this parking space in the current vehicle record), the system will still recognize and output the captured data. The specific principle is shown in steps S2331 to S2332.

[0139] In step S2331, if it is a parallel parking space and there is no detour sign, or if it is a perpendicular parking space and there is no vehicle on site, or if it is an angled parking space and there is no vehicle on site, then the current inspection data is accepted; proceed to step S2340 to end the process.

[0140] In step S2332, if it is a parallel parking space with a detour sign, or if it is a perpendicular parking space with a vehicle in the parking area, or if it is an angled parking space with a vehicle in the parking area, then determine whether the license plate number of the target vehicle is in the list of passing vehicles.

[0141] Step S2332-1: If there is a license plate number of the target vehicle, modify the current inspection data according to the license plate recognition data of the target vehicle; proceed to step S2340 to end the process.

[0142] In step S2332-2, if there is no license plate number for the target vehicle, the current inspection data is invalidated. Proceed to step S2340 to end the process.

[0143] Step S2340: End the process.

[0144] In summary, see Figure 3 The data cleaning steps when the berth is vacant are as follows: If the current inspection data has the additional information for filter vehicle B, then determine whether filter vehicle B is currently present and has an presence confirmation mark. If not, then accept and output the current inspection data. If so, then continue to determine whether the berth number recorded by filter vehicle B is consistent with the berth number in the berth information. If they are consistent, then correct the license plate of the current inspection data to the license plate of the currently present vehicle, change the berth status to occupied, and output the result. If they are inconsistent, then accept and output the current inspection data.

[0145] If the current inspection data does not filter for vehicle B, then the parking space type is determined. If it is a parallel parking space and no detour is taken, the current inspection data is accepted and output. If it is a parallel parking space and a detour is taken, or a perpendicular parking space, or an angled parking space, then it is determined whether the target vehicle A is in the target parking space. If not, the current inspection data is accepted and output. If it is, then it is determined whether the passing vehicle list in the additional information contains the target vehicle A. If it is, then the current inspection data is determined to be "detour unknown" and the current inspection data is invalidated. If it is not, then the license plate of the vehicle in the current inspection data is corrected to the license plate of the currently present vehicle A, the parking space status is changed to occupied, and the result is output.

[0146] Based on the above analysis and judgment, the corrected data of the parking space vacancy capture is finally output for subsequent parking trajectory fusion.

[0147] Step S3000: Parking trajectory fusion is performed based on the cleaned data to obtain a parking trajectory that can reconstruct the actual parking situation. Further, see... Figure 4 Step S3000 includes steps S3100 to S3400, through which the cleaned and automatically corrected captured data are fused into the parking trajectory of the vehicle in the parking space. The cleaned results output in the aforementioned steps include parking space occupancy / vacancy, license plate information, capture time, captured image, license plate image, etc.

[0148] Step S3100: Determine whether the parking space is occupied based on the parking space status in the cleaned data. If the parking space status in the captured data is occupied and there is no vehicle record for that parking space in the system, then use the vehicle information, parking space information, and capture time of this capture data as the entry time to generate an on-site record.

[0149] In step S3200, if the parking space status is occupied, determine whether there is a parking record based on the list of vehicles in the parking space to determine whether the target parking space is occupied.

[0150] Step S3210: If there is a parking record, determine whether the license plate number in the cleaned data matches the license plate number of the vehicle parked at the target parking space. If the parking space status in the captured data is occupied and there is already a vehicle record for that parking space in the system, determine whether the parking sign of the target parking space matches the license plate of the vehicle captured in this instance.

[0151] Step S3211: If the match is consistent, confirm the presence of the vehicle parked at the target parking space; proceed to step S3400.

[0152] Step S3212: If there is no consistency, move the target parking space to the exit parking area; proceed to step S3220.

[0153] In step S3220, if there is no parking record, the vehicle will enter the parking space based on the cleaned data; that is, the license plate will enter the parking space at the time of the capture, and then proceed to step S3400.

[0154] Step S3300: If the parking space status is unoccupied, determine whether the target parking space is occupied; Step S3310: If it is occupied, remove the occupied parking space from the target parking space and proceed to step S3400; If it is not occupied, proceed to step S3400.

[0155] Step S3400: Output the parking trajectory and end.

[0156] In this embodiment, step S3000 performs parking trajectory fusion based on the periodically collected and corrected captured data (determining how long a vehicle stayed in a parking space, and using the collected parking data to confirm the parking record during the stay, recording the time and corresponding captured image). This allows the final parking trajectory to reconstruct the parking facts and become strong evidence, reducing the rate of complaints from car owners, increasing the operator's revenue, and reducing the possibility of public opinion incidents related to smart parking. Through step S3000, a complete parking trajectory information is output, allowing users to view the parked vehicle information, parking space number, entry and exit images, and entry and exit times. The periodically collected parking captured images during the parking process can also be displayed in a timeline format.

[0157] In summary, the method of this embodiment can effectively improve the overall efficiency and accuracy of the parking management system, reduce the rate of manual intervention, make the parking management system more intelligent and cost-effective, support more refined parking management, and enhance user experience.

[0158] See Figure 5 This invention provides a data cleaning and fusion device for roadside parking trajectories, applied to a server, comprising:

[0159] The data acquisition module 1000 is used to acquire the current inspection data. The current inspection data is obtained by the inspection vehicle capturing images of the target parking spaces. The inspection vehicle captures images of all parking spaces in the parking area once within the current capture cycle. The current inspection data includes license plate recognition data, parking space status, and additional information. The additional information includes detour signs, filtered license plates, and a list of passing vehicles. The parking space status includes the parking space number, whether it is occupied or available.

[0160] The judgment and cleaning module 2000 is used to clean the data based on the current inspection data and obtain the cleaned data.

[0161] The trajectory fusion module 3000 is used to fuse parking trajectories based on the cleaned data to obtain parking trajectories that can reconstruct the actual parking situation.

[0162] The judgment and cleaning module 2000 includes:

[0163] The first berth occupancy determination module 2100 is used to determine whether the target berth is occupied based on the berth status.

[0164] The first berth occupancy module 2200 is used to execute the first on-site vehicle judgment module if the target berth is occupied.

[0165] The first vehicle presence determination module 2220 is used to determine whether there is a vehicle with the same license plate based on license plate recognition data and the vehicle presence list; wherein, the vehicle presence list is stored in the server and is a record table of vehicle entry in the parking area during the current capture period; including:

[0166] The "No Vehicle on Site, No Detour Module 2230" is used to recognize the current inspection data if there is no vehicle on site with the same license plate and no detour sign.

[0167] The vehicle detour module 2240 is used to perform data cleaning by combining whether there is a vehicle on the target parking space, additional information, and license plate recognition data if there is no vehicle on the target parking space with the same license plate and a detour sign.

[0168] There is a vehicle module 2250, which is used to determine adjacent parking spaces if there are vehicles with the same license plate, and to perform data cleaning based on the results of the adjacent parking space determination.

[0169] The first unoccupied berth module 2300 is used to perform data cleaning based on the current inspection data and the filtered license plate list if the target berth is not occupied.

[0170] The apparatus provided in the embodiments of this application has the same inventive concept as the method provided in the embodiments of this application. As long as the method can solve the technical problem, the apparatus can also solve the technical problem. This will not be elaborated here.

[0171] Reference Figure 6 The present invention also provides an electronic device 400, including a communication interface 401, a processor 402, a memory 403, and a bus 404. The processor 402, the communication interface 401, and the memory 403 are connected through the bus 404. The memory 403 is used to store a computer program that supports the processor 402 in executing the data cleaning and fusion method for roadside parking trajectories. The processor 402 is configured to execute the program stored in the memory 403.

[0172] Optionally, embodiments of the present invention also provide a computer-readable medium having non-volatile program code executable by a processor 402, the program code causing the processor 402 to perform the data cleaning and fusion method for roadside parking trajectories as described in the above embodiments.

[0173] As is known from common technical knowledge, this invention can be implemented through other embodiments that do not depart from its spirit or essential characteristics. Therefore, the disclosed embodiments described above are merely illustrative in all respects and are not the only ones. All modifications within the scope of this invention or its equivalents are included in this invention.

Claims

1. A method for cleaning and fusing roadside parking trajectory data, characterized in that, Applied to a server, the method includes: Acquire current inspection data; wherein, the current inspection data is obtained by the inspection vehicle capturing images of the target parking space, and the inspection vehicle captures images of all parking spaces in the parking area once within the current capture cycle; the current inspection data includes license plate recognition data, parking space status, and additional information; the additional information includes detour signs, filtered license plates, and a list of passing vehicles; the parking space status includes parking space number, occupied, and vacant; Data cleaning is performed based on the current inspection data to obtain cleaned data, including: Determine whether the target berth is occupied based on the berth status; If the target parking space is occupied, the system determines whether there is a vehicle with the same license plate based on the license plate recognition data and the list of vehicles in the parking area; wherein, the list of vehicles in the parking area is stored in the server and is a record of the entry of vehicles in the parking area during the current capture period. If there is no vehicle with the same license plate on site and it does not have a detour sign, the current inspection data is accepted. If there is no vehicle with the same license plate on site and it carries the detour sign, then data cleaning is performed by combining whether there is a vehicle on site at the target parking space, the additional information, and the license plate recognition data. If there are vehicles with the same license plate on site, then adjacent parking spaces are determined, and data cleaning is performed based on the results of the adjacent parking space determination. If the target parking space is not occupied, data cleaning is performed based on the current inspection data and the filtered license plate list; The parking trajectory is fused based on the cleaned data to obtain a parking trajectory that can reconstruct the actual parking situation.

2. The data cleaning and fusion method for roadside parking trajectories according to claim 1, characterized in that, The current inspection data also includes captured GPS data, GPS signal strength, patrol vehicle riding type, and capture time; wherein, the license plate recognition data includes the target vehicle's license plate number, license plate color, and recognition reliability, and the patrol vehicle riding type includes normal and detour.

3. The method for cleaning and fusing roadside parking trajectory data according to claim 1, characterized in that, The license plate recognition data also includes license plate images; the data cleaning process, which combines the presence of a vehicle at the target parking space, the additional information, and the license plate recognition data, includes: Determine whether there is a vehicle at the target berth based on the berth number and the list of vehicles on site; If there is a vehicle in the target parking space, data cleaning is performed based on the list of passing vehicles and the license plate image; If there is no vehicle at the target berth, the current inspection data is accepted.

4. The method for cleaning and fusing roadside parking trajectory data according to claim 3, characterized in that, The data cleaning based on the list of passing vehicles and the license plate images includes: Determine whether the list of passing vehicles contains the license plate information of the vehicle at the target parking space; If the license plate information of the vehicle in the target parking space is included, the license plate recognition data is modified according to the license plate information of the vehicle in the target parking space. If the license plate information of the vehicle in the target parking space is not included, then determine whether the number of pixels in the license plate image is greater than a preset pixel threshold. If the value exceeds a preset pixel threshold, the current inspection data is discarded. If the value is not greater than a preset pixel threshold, the current inspection data is accepted.

5. The method for cleaning and fusing roadside parking trajectory data according to claim 2, characterized in that, The process of determining adjacent berths and cleaning the data based on the results includes: Determine whether the berth number of the target berth stored in the server and the berth number of the target berth captured in this snapshot are adjacent berth numbers; If the parking space numbers are adjacent, the on-site confirmation record of the target vehicle is queried, and the data is cleaned based on the query results. If it is not an adjacent berth number, then the current inspection data is accepted.

6. The method for cleaning and fusing roadside parking trajectory data according to claim 5, characterized in that, The process of querying the presence confirmation records of the target vehicle and cleaning the data based on the query results includes: Determine whether the target vehicle has a record of being present based on its license plate number and the list of vehicles present; If there is a record of presence, the parking space number in the parking space status will be modified according to the parking space number of the target vehicle. Also, if there is uncleaned data in the current capture cycle and the corresponding data of the target vehicle is still stored in the parking vehicle list, the corresponding data of the target vehicle stored in the parking vehicle list will be invalidated. If there is no record of presence, the current inspection data is accepted.

7. The method for cleaning and fusing roadside parking trajectory data according to claim 1, characterized in that, The additional information also includes berth type, which includes parallel berths, perpendicular berths, and angled berths; if the target berth is not occupied, data cleaning is performed based on the current inspection data and the filtered license plate list, including: Determine if the filtered license plate information for the target vehicle is present in the filtered license plate list; If there is license plate information corresponding to the target filtering vehicle, then it is determined whether the target filtering vehicle is an on-site vehicle and has an on-site confirmation mark based on the license plate information. The on-site confirmation mark is obtained from the on-site vehicle record and is used to indicate that the license plate of the target filtering vehicle has been captured multiple times at the target parking space. If the vehicle is present and has an presence confirmation mark, then determine whether the parking space number corresponding to the filtered license plate information is consistent with the parking space number in the parking space status. If they match, the license plate recognition data is modified according to the filtered license plate information, and the parking space status is changed to occupied. If there is a discrepancy, the current inspection data is accepted. If the vehicle is not present, then the current inspection data is accepted. If there is no license plate information corresponding to the target vehicle, then determine the parking space type; If it is a parallel berth and there is no detour sign, or if it is a perpendicular berth and there is no vehicle on site, or if it is an angled berth and there is no vehicle on site, then the current inspection data is accepted. If it is a parallel parking space with the detour sign, or if it is a perpendicular parking space with a vehicle in the parking space, or if it is an angled parking space with a vehicle in the parking space, then determine whether the target vehicle's license plate number is in the list of passing vehicles. If the license plate number of the target vehicle is available, the current inspection data is modified based on the license plate recognition data of the target vehicle. If the target vehicle's license plate number is not available, the current inspection data will be invalidated.

8. The method for cleaning and fusing roadside parking trajectory data according to claim 1, characterized in that, The step of fusing parking trajectories based on the cleaned data to obtain parking trajectories that can reconstruct the actual parking situation includes: Based on the berth status in the cleaned data, determine whether it is occupied; If the parking space is occupied, check the list of vehicles in the parking area to determine if there is a parking record. If there is a parking record, determine whether the license plate number in the cleaned data matches the license plate number of the vehicle parked in the target parking space; If they match, confirm the presence of the vehicle parked at the target parking space; then proceed to the step of outputting the parking trajectory. If there is a discrepancy, the parking space at the target parking spot will be removed, and the parking space will be re-entered based on the data after cleaning. If there is no parking record, then entry is performed based on the cleaned data; proceed to the step of outputting the parking trajectory; If the parking space is unoccupied, then determine whether the target parking space is occupied. If there is a parking space, the parking space in the target parking space will be exited; proceed to the step of outputting the parking trajectory. If the vehicle is not parked, proceed to the step of outputting the parking trajectory; Output parking trajectory.

9. The method for cleaning and fusing roadside parking trajectory data according to claim 1, characterized in that, The step of determining whether there is a vehicle with the same license plate based on the license plate recognition data and the list of vehicles present includes, prior to: If the current inspection data already exists, it is determined based on historical inspection data. If the current inspection data already exists, the berth status in the current inspection data is modified to an empty berth. The historical inspection data is stored in the server and is the inspection data captured by the inspection vehicle in at least one historical capture cycle.

10. A data cleaning and fusion device for roadside parking trajectories, characterized in that, Applied to servers, including: The data acquisition module is used to acquire current inspection data; wherein, the current inspection data is obtained by the inspection vehicle capturing images of the target parking space, and the inspection vehicle captures images of all parking spaces in the parking area once within the current capture cycle; the current inspection data includes license plate recognition data, parking space status, and additional information; the additional information includes detour signs, filtered license plates, and a list of passing vehicles; the parking space status includes parking space number, occupied, and vacant; The judgment and cleaning module is used to clean the data based on the current inspection data to obtain cleaned data; The trajectory fusion module is used to fuse parking trajectories based on the cleaned data to obtain parking trajectories that can reconstruct the actual parking situation. The judgment and cleaning module includes: The first berth occupancy determination module is used to determine whether the target berth is occupied based on the berth status. The first berth occupancy module is used to execute the first on-site vehicle judgment module if the target berth is occupied. The first vehicle presence determination module is used to determine whether there is a vehicle with the same license plate based on the license plate recognition data and the vehicle presence list; wherein, the vehicle presence list is stored in a server, and the vehicle presence list is an entry record table of vehicles in the parking area during the current capture period; including: The "No Vehicles On-Site and No Detour" module is used to recognize the current inspection data if there are no vehicles on-site with the same license plate and no detour markings. The vehicle detour module is used to perform data cleaning by combining whether there is a vehicle in the target parking space, the additional information, and the license plate recognition data if there is no vehicle in the parking space with the same license plate and the vehicle has the detour mark. There is an on-site vehicle module, which is used to determine adjacent parking spaces if there are on-site vehicles with the same license plate, and to perform data cleaning based on the results of the adjacent parking space determination. The first unoccupied berth module is used to perform data cleaning based on the current inspection data and the filtered license plate list if the target berth is not occupied.

Images