KR102173121B1 - System for confirming parking space based on image and method thereof - Google Patents

Abstract

According to the present invention, disclosed are a parking space checking system based on an image and a method thereof. In other words, the system of the present invention is capable of accurately recognizing the shape of a vehicle in collected data, by collecting unstructured data for AI learning, processing the collected unstructured data for AI learning by segmentation, training a plurality of AI model architecture candidates through the unstructured data for AI learning processed by the segmentation, measuring the accuracy for each of the plurality of AI model architecture candidates using a performance measurement index, and selecting the optimal AI model based on the measured accuracy for each of the plurality of AI model architecture candidates.

Description

System for confirming parking space based on image and method thereof

The present invention relates to an image-based parking space identification system and method thereof, in particular, collecting unstructured data for AI learning, segmentation processing the collected unstructured data for AI learning, and processing the unstructured AI learning unstructured data subjected to the segmentation processing. Through the training, multiple AI model architecture candidates are trained, the accuracy of multiple AI models is measured using performance metrics, and the optimal AI model is selected based on the measured accuracy of multiple AI models. It relates to a parking space identification system and a method thereof.

Dissatisfaction with drivers regarding parking problems in Korea has been rising since long ago.

In addition, in order to solve various parking problems, information such as parking lot location, hourly price, and parking lot type is provided, but there is an inconvenience that users have to go directly to the parking lot because there is no detailed information such as parking space status.

In addition, in order to provide the status of parking spaces in the parking lot, enormous cost problems arise as public institutions and private businesses that operate parking lots purchase and operate expensive equipment, or consider additional labor costs due to the operation of expensive equipment. .

Korean Patent Laid-Open Patent No. 10-2019-0084903 [Title: Parking lot sharing method and device for efficient use of parking space]

An object of the present invention is to collect unstructured data for AI learning, segmentation processing the collected unstructured data for AI learning, train a plurality of AI model architecture candidates through the segmentation-processed unstructured data for AI learning, and perform performance measurement indexes. It is to provide a parking space identification system and method based on an image that measures the accuracy of a plurality of AI models by using and selects an optimal AI model based on the accuracy of the measured plurality of AI models.

Another object of the present invention is to process image information acquired through a camera installed in a parking lot, and apply an AI model selected prior to the processed image information to determine a parking space and a non-parking space within the image information. It is to provide a system and method for checking a parking space based on an image that recognizes and displays an available parking space and a non-parking space within the recognized image information.

A system for checking a parking space based on an image according to an exemplary embodiment of the present invention includes: a parking management server for obtaining image information related to a parking lot under management and transmitting the obtained image information; And performing preprocessing on the image information, applying the preprocessed data to a preselected AI model, generating an output value according to AI modeling, and a plurality of parking lots related to the parking lot management server stored in advance in the server. Calculate the information on the parking space and the performance measurement index of the output value according to the AI modeling, determine whether the calculated at least one performance measurement index is equal to or greater than a preset threshold, and as a result of the determination, the calculated at least one performance When there is at least one performance measurement index that is equal to or greater than the threshold value among the measurement indexes, it is determined that at least one parking space related to the at least one performance measurement index can be parked, and as a result of the determination, the calculated at least one performance When there is at least one other performance measurement index that is less than the threshold value among the measurement indexes, it is determined that at least one other parking space related to the other at least one performance measurement index is not allowed to park, and the parking lot according to the determination result Including the server for generating parking status information including a parking space available and non-parking parking spaces among a plurality of parking surfaces related to a parking lot managed by the management server, and transmitting the generated parking status information to the parking management server can do.

As an example related to the present invention, the output value may include information indicating that a vehicle is included or information indicating that a vehicle is not included for each of the preprocessed data.

As an example related to the present invention, the parking status information includes date and time information at the time of generating the parking status information, parking lot name, location information of the parking lot, contact information of the parking lot, parking image, total number of parking spaces, number of parking spaces available for parking. , At least one of the number of non-parking parking spaces, location information on the available parking spaces, and location information on the non-parking parking spaces.

As an example related to the present invention, the server, in relation to a plurality of parking lots managed by the server, collects vehicle flow data for each time slot for each of the plurality of parking lots, and vehicle flow data for each time slot for the collected parking lots. Based on the calculation of the probability of parking available for each time slot for each of a plurality of parking lots, and when receiving the destination and expected arrival date and time transmitted from the terminal, the probability of parking available for each time slot for the plurality of parking lots managed by the server, a plurality of Based on the location information of the parking lot, the received destination, and the scheduled arrival date and time, the terminal searches for a specific parking lot with the highest possible parking possibility at the time of reaching the destination, and relates to the terminal and the searched specific parking lot. In conjunction with the parking lot management server, a parking reservation function may be performed for a specific parking lot with the highest probability of the searched parking available.

According to an embodiment of the present invention, a method for checking a parking space based on an image includes performing, by a server, pre-processing on the image information transmitted from the parking management server; Generating an output value according to AI modeling by applying the preprocessed data to a pre-selected AI model by the server; Calculating, by the server, information on a plurality of parking spaces related to a parking lot corresponding to the parking lot management server stored in advance in the server and a performance measurement index of an output value according to the AI modeling; Determining, by the server, whether the calculated at least one performance measurement index is equal to or greater than a preset threshold; As a result of the determination, when there is at least one performance measurement index that is equal to or greater than the threshold value among the calculated one or more performance measurement indexes, the server may park at least one parking space related to the at least one performance measurement index. Determining the presence state; As a result of the determination, when there is at least one other performance measurement index less than the threshold value among the calculated one or more performance measurement indicators, the server determines at least one other parking space related to the other at least one performance measurement index. Determining that parking is not possible; Generating, by the server, parking status information including parking spaces that can be parked and parking spaces that cannot be parked from among a plurality of parking surfaces related to a parking lot managed by the parking management server according to the determination result; And transmitting, by the server, the generated parking status information to the parking lot management server.

The pre-processing of the image information according to an embodiment of the present invention may include: extracting a specific image from image information received in real time according to a preset analysis unit for a plurality of images constituting the image information; And converting the extracted pixels of the specific image into an array or separating channels for a plurality of pixels constituting the specific image.

Collecting, by the server according to an embodiment of the present invention, vehicle flow data for each time slot for each of the plurality of parking lots in relation to a plurality of parking lots managed by the server; Calculating, by the server, a parking availability probability for each time slot for each of the plurality of parking lots based on the collected vehicle flow data for each time slot for the plurality of parking lots; When receiving the destination and the expected arrival date and time transmitted from the terminal, the server, the probability of parking available for each time slot for a plurality of parking lots managed by the server, location information of the plurality of parking lots, the received destination and the Searching for a specific parking lot having the highest parking availability probability at a time point at which the terminal reaches the destination based on a scheduled arrival date and time; And performing, by the server, a parking reservation function for the specific parking lot having the highest possible parking availability probability by linking with the terminal and a parking management server related to the searched specific parking lot.

The step of collecting vehicle flow data for each time slot for each of the plurality of parking lots according to an embodiment of the present invention includes accumulating a plurality of parking status information for each parking lot generated from a plurality of image information related to a plurality of parking lots received in real time. process; And collecting vehicle flow data for each time slot related to the parking lot based on the accumulated parking status information for each parking lot.

The vehicle flow amount data for each time period according to an embodiment of the present invention may include at least one of a parking lot name, location information of a parking lot, a total number of parking spaces, and the number of parking spaces available for parking.

Transmitting, by the server according to an embodiment of the present invention, encrypted information generated by encrypting vehicle information and parking reservation function result information using a public key set in advance to a parking management server performing a reservation function; Decrypting, by the parking lot management server, vehicle information and parking reservation function result information included in the encryption information using a private key managed by the parking lot management server; And when the vehicle reserved by the user of the terminal attempts to enter the parking lot managed by the parking management server, the parking reserved by the terminal based on the vehicle information and the parking reservation function result information by the parking management server. It may further include performing at least one of an entry/exit management function, an open/close management function of a circuit breaker at a parking lot entrance, and a simple parking fee settlement function for the available space.

The present invention collects unstructured data for AI learning, performs segmentation processing of the collected unstructured data for AI learning, trains a plurality of AI model architecture candidates through the segmentation-processed unstructured data for AI learning, and uses a performance measurement index. By measuring the accuracy of a plurality of AI models and selecting an optimal AI model based on the measured accuracy of a plurality of AI models, there is an effect of accurately recognizing the shape of the vehicle in the collected data.

In addition, the present invention processes the image information acquired through the camera installed in the parking lot, and recognizes the available parking space and the non-parking space within the image information by applying the AI model selected before the processed image information. And, by displaying available parking spaces and non-parking spaces within the recognized image information, the current status of the entire parking lot can be grasped only with the existing CCTV equipment without the input of expensive equipment or manpower required to manage the parking status in the parking lot, It has the effect of improving the efficiency of parking lot management by expanding it to a 24-hour unmanned system.

1 is a block diagram showing the configuration of a system for checking a parking space based on an image according to an embodiment of the present invention.
2 is a diagram showing the definition of a performance measurement index according to an embodiment of the present invention.
3 to 4 are flowcharts illustrating a method of checking a parking space based on an image according to an exemplary embodiment of the present invention.
5 is a diagram illustrating an example of an image extracted from image information according to an embodiment of the present invention.
6 is a diagram showing an example of an output value to which an AI model according to an embodiment of the present invention is applied.

It should be noted that the technical terms used in the present invention are used only to describe specific embodiments, and are not intended to limit the present invention. In addition, the technical terms used in the present invention should be interpreted as generally understood by those of ordinary skill in the technical field to which the present invention belongs, unless otherwise defined in the present invention, and is excessively comprehensive. It should not be construed as a human meaning or an excessively reduced meaning. In addition, when a technical term used in the present invention is an incorrect technical term that does not accurately express the spirit of the present invention, it should be replaced with a technical term that can be correctly understood by those skilled in the art. In addition, general terms used in the present invention should be interpreted as defined in the dictionary or according to the context before and after, and should not be interpreted as an excessively reduced meaning.

In addition, the singular expression used in the present invention includes a plurality of expressions unless the context clearly indicates otherwise. In the present invention, terms such as “consisting of” or “comprising” should not be construed as necessarily including all of the various components or steps described in the invention, and some components or some steps may not It should be construed that it may or may further include additional components or steps.

In addition, terms including ordinal numbers such as first and second used in the present invention may be used to describe the constituent elements, but the constituent elements should not be limited by the terms. The terms are used only for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, a first component may be referred to as a second component, and similarly, a second component may be referred to as a first component.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings, but the same or similar components are assigned the same reference numerals regardless of the reference numerals, and redundant descriptions thereof will be omitted.

In addition, in describing the present invention, when it is determined that a detailed description of a related known technology may obscure the subject matter of the present invention, a detailed description thereof will be omitted. In addition, it should be noted that the accompanying drawings are only for easily understanding the spirit of the present invention and should not be construed as limiting the spirit of the present invention by the accompanying drawings.

1 is a block diagram showing the configuration of a system 10 for checking a parking space based on an image according to an exemplary embodiment of the present invention.

As shown in FIG. 1, the system 10 for checking a parking space based on an image includes a server 100 and a parking management server 200. Not all components of the parking space identification system 10 shown in FIG. 1 are essential components, and the parking space identification system 10 may be implemented by more components than those shown in FIG. 1, The parking space identification system 10 may be implemented with fewer components.

Each of the server 100 and the parking lot management server 200 is a communication unit (not shown) for performing communication functions with other terminals, and a storage unit (not shown) for storing various information and programs (or applications) , A display unit (not shown) for displaying various information and program execution results, an audio output unit (not shown) for outputting audio information corresponding to the various information and program execution results, and various components and functions of each terminal. It may include a control unit (not shown) for controlling.

The server 100 includes a smart phone, a portable terminal, a mobile terminal, a foldable terminal, a personal digital assistant (PDA), and a portable multimedia device (PMP). Player) Terminal, Telematics Terminal, Navigation Terminal, Personal Computer, Notebook Computer, Slate PC, Tablet PC, Ultrabook, Wearable Device (Wearable) Device, for example, including smartwatch, smart glass, HMD (Head Mounted Display), Wibro terminal, Internet Protocol Television (IPTV) terminal, smart TV, digital broadcasting It can be applied to various terminals such as terminals, audio video navigation (AVN) terminals, audio/video (A/V) systems, flexible terminals, and digital signage devices.

In addition, the server 100 may be applied to various servers such as a web server, a database server, a proxy server, and a distributed server.

In addition, the server 100 communicates with one or more parking lot management servers 200 and terminals (not shown).

In addition, the server 100 performs a membership registration procedure for a user (or administrator) such as the parking lot management server 200.

In addition, the server 100 registers (or shares) personal information (or company information) related to a user such as the parking lot management server 200.

In addition, the server 100 performs a member management function for a user such as the parking lot management server 200.

In addition, the server 100 provides a dedicated app and/or web site for performing a subscription function, a parking status information providing function, and the like, to the parking management server 200.

In addition, the server 100 provides a bulletin board function for announcements and events.

In addition, the server 100 interlocks with an arbitrary parking lot management server 200 to set a subscription function in relation to a parking space provision service provided by the server 100.

That is, a plurality of input values corresponding to a plurality of input items according to a user input (or user selection/touch/control) of the corresponding parking management server 200 in the subscription application screen displayed on the parking management server 200 are included. When receiving the subscription request information, the server 100 receives subscription application information transmitted from the parking lot management server 200, identification information of the parking lot management server 200, and the like.

In addition, the server 100 sets a subscription to the parking management server 200 registered with the server 100 with the corresponding registration application information according to registration application information and subscription conditions included in the received subscription application information.

At this time, when the subscription condition is paid, the server 100 interlocks with the parking management server 200 and a payment server (not shown), and a corresponding parking management server 200 in which a subscription function is set according to the subscription application information. ) Related to the monthly (or case-by-case/weekly) periodic payment function (e.g., card payment, bank payment account) for the parking status information generated based on the video information provided by the corresponding parking lot management server 200 Automatic transfer through linkage, payment using cash points or cash remaining in the account of the parking lot management server 200 registered as a member of the server 100, simple payment including Kakao Pay, Naver Pay, etc.) Do it automatically. At this time, when the subscription period fee is set as the total fee according to the subscription period, the server 100 may perform a payment function for the total fee by interworking with the parking lot management server 200 and the payment server. have. In addition, the server 100 may provide differentiated additional services (eg, an additional discount function, an automatic subscription period extension function, etc.) differentiated according to the membership level for each monthly/yearly price section.

In addition, the server 100 provides a payment function execution result to the parking lot management server 200. Here, the payment function execution result includes unique identification information (or unique identification number/member unique number), subscription conditions, subscription period, subscription start date, e-mail address, payment amount information, payment date and time information, and the like.

In addition, the server 100 receives image information transmitted (or provided) from the parking lot management server 200. Here, the image information is image information related to a parking lot managed by the parking lot management server 200, is obtained through a camera (not shown) installed at a fixed location in the parking lot, and is divided into a plurality of (or One or more) parking surfaces, and one or more cars parked on the parking surfaces. In this case, the server 100 may receive image information continuously transmitted from the parking lot management server 200 in a streaming format. Here, the parking management server 200 may be in a state in which a subscription function is set in relation to a parking space providing service provided by the server 100.

In addition, the server 100 performs preprocessing on the received image information (or streaming image information).

That is, the server 100 extracts a specific image (or a specific frame) from image information received in real time according to a preset analysis unit for a plurality of images (or a plurality of frames) constituting the received image information, and , Preprocessing is performed on the extracted specific image. Here, the pre-processing may be a function of converting pixels of the specific image (or image information) into an array, or a function of separating channels (eg, separating into RGB channels) for a plurality of pixels constituting the specific image. have.

In this way, the server 100 samples (or extracts) a specific image at a preset period (or analysis unit) among image information received in real time in order to reduce the load due to preprocessing, etc., and the sampled Pre-processing can be performed on a specific (or extracted) image.

In an embodiment of the present invention, it is described that the server 100 extracts a specific image according to the analysis unit from among a plurality of images constituting the received image information, and performs preprocessing on the extracted specific image. However, the present invention is not limited thereto, and the server 100 may receive a specific image from the parking lot management server 200 according to the analysis unit, and may perform preprocessing on the received specific image.

In addition, the server 100 generates (or outputs) an output value according to AI modeling by applying the preprocessed data to a preselected (or set) AI model. Here, the AI model includes EfficientNet-L2, DeepLabv3+, and the like.

That is, the server 100 processes the preprocessed data as an input value of the AI model and applies it to the AI model, and then generates (or outputs) an output value according to the application of the AI model. Here, the output value is information indicating that a vehicle is included (for example, a preset '1') or information indicating that a vehicle is not included (for example, for each of the preprocessed data (or for each pixel in the preprocessed image)) For example, it includes a preset '0' or an inverse value of information indicating that the vehicle is included).

In this case, the AI model may be in a state selected through the following process.

That is, the server 100 accesses (or works in conjunction with) a plurality of content management servers (not shown) that provide various types of content, and collects vehicle images through a crawling method. Here, the content management server may be a parking lot CCTV management server, a social network service (SNS) server, a video content providing server, a cloud server, etc., and Instagram, Facebook, Twitter ( Twitter), YouTube, and other servers.

In addition, in order to collect only vehicle images in which the shape of the vehicle is well displayed, the server 100 considers the location of the parking lot CCTV among the collected one or more (or a plurality of) vehicle images, and is 45 degrees from the ground at a certain height. It is possible to filter (or collect) car images taken at an angle of ~ 90 degrees (or perpendicular to the horizontal plane).

In this way, the server 100 may collect vehicle images, which are unstructured data for AI learning taken at a high angle so that the shape of the vehicle can be well confirmed.

In addition, the server 100 performs segmentation processing on the collected one or more automobile images (or unstructured data).

That is, the server 100 labels (or segmentation) whether a pixel of the collected vehicle image is a pixel with a vehicle (1) or a pixel without a vehicle (0) as 1 or 0.

As described above, when learning an object through an image segmentation model, the server 100 restricts the target object to a vehicle (or vehicle) and reduces the learning space by binary classification. Accordingly, learning efficiency can be improved.

In addition, the server 100 trains a plurality of pre-selected AI model architecture candidates using the processed vehicle image (or processed data set). Here, the plurality of AI model architecture candidates use a deep neural network and a model that mainly performs convolution compute is selected, EfficientNet-L2, DeepLabv3+, DPC, and CFNet. , DCNAS, DeepLabv3-JFT, CASIA_IVA_SDN, Smooth Network with Channel Attention Block, SANet, etc.

In addition, the server 100 measures accuracy and derives (or determines/selects/selects) an optimal model for each of a plurality of AI model architecture candidates according to the training using the performance measurement index.

That is, as shown in Fig. 2, the server 100 calculates the correct answer segmentation (A) and the predicted segmentation (B) of the AI model architecture candidate using the definition of the performance measurement index, and the AI according to the training Accuracy is measured (or calculated/calculated) for each model architecture candidate. In this case, as shown in FIG. 2, the more the two segmentations overlap, the closer to 1, and the less overlapping, the closer to 0.

In addition, the server 100 is EfficientNet-L2, which shows the highest accuracy (for example, 75.2%) among the measured AI model architecture candidate-specific accuracy, and DeepLabv3+, which shows the next high accuracy (for example, 74.3%). Is selected (or selected) as the optimal AI model. At this time, the configuration of the AI model for training (or test) is EfficientNet-L2, Backbone is EfficientNet-L2, Pretrained with ImageNet, Finetuned with PASCAL VOC 2012, and DeepLabv3+, Backbone is Xception. , Pretrained with ImageNet, and Finetuned with PASCAL VOC 2012.

The server 100 may select an AI model applied in an embodiment of the present invention through such a process.

In addition, the server 100 recognizes a parking space related to a specific parking lot by the following process.

That is, the server 100 sorts the space occupied by a large number of cars among the total predictions of the AI model in descending order. Here, the space occupied by the automobile may mean that, for example, a weight is assigned to each pixel in the 1920×1080 image, and the positions of the pixels with the highest number of 1 are sorted in descending order.

For example, the server 100 sorts, in descending order, information on whether a car accumulated for each pixel has ever been parked among all predictions of the AI model accumulated for more than one day.

In addition, the server 100 is as much as the number of parking spaces set in advance in relation to the specific parking lot among the spaces occupied by a large number of cars among the total prediction of the sorted AI model (or a preset ratio in the corresponding parking space-for example, 80%). Select (or separate) the upper space (or upper pixel) as many as the number corresponding to -. Here, the method of separating the upper pixel (or upper space) is to box the parking space according to the administrator's selection, measure the number of parking space pixels relative to the total number of image pixels, extract the upper pixel by the measured number, and park. Represents processing as spatial pixels.

For example, the server 100 selects a space above a preset number of parking spaces (for example, 30) from information on whether or not a vehicle accumulated for each of the sorted pixels has been parked.

In addition, the server 100 uses the selected (or separated) upper space (or upper pixel/parking space information) to calculate a performance measurement index with an output value according to AI modeling that is output in real time. You can determine whether or not parking is available.

In addition, the server 100 includes information on a plurality of parking spaces related to a parking lot corresponding to the parking management server 200 stored in advance (or registered) in the server 100 and the performance of the output value according to the AI modeling. Calculate (or calculate) metrics. At this time, the information on the parking space is location information on one or more parking spaces in relation to the parking lot corresponding to the parking lot management server 200, a labeling value (or unique identification information/number) corresponding to the location information, and parking It includes the total number of spaces (or total number of parking spaces/total number of parking spaces). In addition, the performance measurement indicators are IoU (Intersection over Union), precision, and reproducibility based on TP (True Positive), FP (False Positive), TN (True Negative), FN (False Negative), etc. Recall), accuracy, etc.

In addition, the server 100 is the calculated (or calculated) one or more performance measurement index (or the calculated information on the stored parking space and the performance measurement index of the output value according to the AI modeling) is a preset threshold Determine (or confirm) whether it is abnormal.

When there is at least one performance measurement index that is equal to or greater than the threshold value among the determination result (or the confirmation result) and the calculated one or more performance measurement indexes, the server 100 may perform at least one performance measurement index related to the at least one performance measurement index. Determining one parking space (or at least one parking space corresponding to information on the parking space related to the at least one performance measurement index that is equal to or greater than the threshold value among the calculated one or more performance measurement indicators) as parking available (or Processing).

In addition, the determination result (or the confirmation result), when there is at least one other performance measurement index less than the threshold value among the calculated one or more performance measurement indicators (or the threshold value among the calculated one or more performance measurement indicators) If there are other performance measurement indicators other than at least one performance measurement index that is the above), the server 100 measures at least one other parking space (or the calculated one or more performance measurements) related to the other at least one performance measurement index. Among the indicators, at least one other parking space corresponding to information on the parking space related to the other at least one performance measurement index that is less than the threshold value) is determined (or processed) as a state in which parking is impossible.

In addition, the server 100 is based on the determination result, the parking status including a parking space (or parking space) and non-parking parking spaces among a plurality of parking surfaces related to the parking lot managed by the parking management server 200 Generate information. Here, the parking status information is the date and time information at the time of creation of the parking status information, the name of the parking lot, the location information of the parking lot, the contact information of the parking lot, the parking lot image (or the parking lot cross section/the above image information/image), the total number of parking spaces, It includes the number of parking spaces available for parking, the number of parking spaces that cannot be parked, location information (or labeling value/unique identification information/number) for parking spaces that can be parked, and location information for parking spaces that cannot be parked.

In addition, the server 100 transmits the generated parking status information to the parking management server 200.

In addition, in relation to a plurality of parking lots managed by the server 100, the server 100 collects vehicle flow data for each time slot for each of the plurality of parking lots. Here, the vehicle flow data for each time period includes the name of the parking lot, location information of the parking lot, the total number of parking spaces, and the number of parking spaces available for parking.

That is, the server 100 accumulates a plurality of parking status information for each parking lot that is generated from a plurality of image information related to a plurality of parking lots received in real time, and corresponds to the accumulated parking status information for each parking lot. Collects vehicle flow data for each parking lot (or vehicle flow data for each parking lot).

In addition, the server 100 calculates (or calculates) a parking availability probability for each of a plurality of parking lots based on the collected vehicle flow data for each time slot for the plurality of parking lots.

That is, the server 100 calculates the parking availability probability for each time slot, which is the probability of parking availability for each parking lot, based on the collected vehicle flow data for each of the plurality of parking lots.

In addition, the server 100 stores the calculated (or calculated) parking availability probability for each time slot for each of the plurality of parking lots. In this case, the server 100 may convert and store the calculated (or calculated) parking availability probability for each time slot for each of the plurality of parking lots.

In addition, the server 100 receives a destination, a scheduled arrival date and time, and identification information of a terminal transmitted from a terminal (not shown).

In addition, the server 100 includes the probability of parking availability by time slot for a plurality of parking lots managed by the server 100, location information of the plurality of parking lots (or vehicle flow data by time slot), the received destination, and arrival schedule. Based on the date and time, the terminal searches (or confirms) a specific parking lot (or a specific parking space in a specific parking lot) with the highest parking probability at the time (or time) when reaching the destination.

At this time, the server 100 transmits the destination, the scheduled arrival date and time, the identification information of the terminal, vehicle information (eg, vehicle type, vehicle width, vehicle height, vehicle length, etc.) transmitted from the terminal. When receiving, the server 100 includes the probability of parking available for each time slot for a plurality of parking lots managed by the server 100, location information of the plurality of parking lots, the received destination, expected arrival date and time, and vehicle information. In consideration of together, it is also possible to search for a specific parking lot (or a specific parking space in a specific parking lot) with the highest possible parking probability at the time when the vehicle reserved (or on board) by the user with the corresponding terminal reaches the destination. Here, the identification information of the terminal includes MDN (Mobile Directory Number), mobile IP, mobile MAC, subscriber identity module (SIM) card unique information, serial number, and the like.

In addition, the server 100 performs a parking reservation function for a specific parking lot with the highest probability of the searched parking available in connection with the terminal and the parking management server 200 related to the searched specific parking lot.

At this time, when performing a parking reservation function for the searched parking lot with the highest probability of parking, the server 100 generates encryption information by encrypting vehicle information and result information of the parking reservation function using a preset public key. . Here, the parking reservation function result information is date and time information at the time of reservation, parking reservation date and time information, reserved parking space information, reserved person information (eg, name, phone number, terminal identification information, etc.), simple payment It includes payment information (including, for example, Kakao Pay, Naver Pay, etc.).

In addition, the server 100 transmits the generated encryption information to the parking lot management server 200 that has performed a reservation function.

The parking lot management server 200 communicates with the server 100 and the terminal.

In addition, the parking lot management server 200 is a user (or administrator) to receive various functions, etc. through a dedicated app and/or website provided by the server 100 by interworking with the server 100 Sign up as a member, and register personal information and the like in the server 100. In this case, the personal information (or company information) includes ID, email address, password (or password), name, gender, date of birth, contact information, address, business name, business registration number, and the like.

In addition, the parking lot management server 200 may subscribe as a user to the server 100 by using SNS account information or other site account information to which a user of the corresponding parking management server 200 has subscribed. Here, the SNS account may be information related to Facebook, Twitter, Instagram, Kakao Story, and Naver Blog. In addition, the other site account (or mobile messenger account) may be information related to YouTube, Kakao, Naver, Kakao Talk, Line, Viber, WeChat, WhatsApp, Telegram, Snapchat, and the like.

In addition, when performing the membership registration procedure, the parking lot management server 200 must complete the authentication function through an identity authentication means (eg, including mobile phone, credit card, i-PIN, etc.) Can be completed normally.

In addition, after membership registration is complete, the parking lot management server 200 is a dedicated app (or application/application/specific app) provided from the server 100 in order to use the service provided by the server 100 Is installed in the corresponding other terminal 200. In this case, the dedicated app may be an app for performing a subscription function, a parking status information providing function, and the like.

In addition, the parking lot management server 200 interworks with the server 100 to set a subscription function in relation to a parking space providing service provided by the server 100.

In other words, the parking lot management server 200 responds to a plurality of input items according to a user input (or user selection/touch/control) of the corresponding parking lot management server 200 on the subscription application screen displayed on the parking lot management server 200. Subscription application information including a plurality of corresponding input values is received.

In addition, the parking management server 200 transmits the received subscription application information and identification information of the parking management server 200 to the server 100.

At this time, when the subscription condition is paid, the parking lot management server 200 interlocks with the server 100 and the payment server, and a subscription function is set according to the subscription application information. For the parking status information generated based on the video information provided by the corresponding parking lot management server 200, a monthly (or case-by-case/weekly) regular payment function (for example, card payment, bank payment account linkage) during the subscription period. Automatic transfer, payment using cash points or cash remaining in the account of the parking lot management server 200 registered as a member of the server 100, including simple payment including Kakao Pay, Naver Pay, etc.) do.

In addition, the parking lot management server 200 receives a payment function execution result transmitted from the server 100 and displays (or outputs) the received payment function execution result. Here, the payment function execution result includes unique identification information (or unique identification number/member unique number), subscription conditions, subscription period, subscription start date, e-mail address, payment amount information, payment date and time information, and the like.

When payment is unsuccessful, the parking lot management server 200 receives information indicating that the payment has failed from the server 100 (eg, insufficient balance, exceeded the limit, etc.), and the received payment has failed. Display (or print out) information indicating status (including, for example, insufficient balance, exceeded limit, etc.).

In addition, the parking lot management server 200 includes one or more video information related to the parking lot (or one included in the parking lot) through one or more cameras (not shown) configured on one side of the parking lot managed by the parking lot management server 200. Image information including the above parking surface) is acquired.

In addition, the parking lot management server 200 transmits the acquired one or more image information to the server 100. In this case, the parking lot management server 200 may continuously transmit the acquired one or more image information to the server 100 in a streaming format.

In addition, the parking lot management server 200 receives parking status information transmitted from the server 100. Here, the parking status information is the date and time information at the time of creation of the parking status information, the name of the parking lot, the location information of the parking lot, the contact information of the parking lot, the parking lot image (or the parking lot cross section/the above image information/image), It includes the number of parking spaces available for parking, the number of parking spaces that cannot be parked, location information (or labeling value/unique identification information/number) for parking spaces that can be parked, and location information for parking spaces that cannot be parked.

In addition, the parking lot management server 200 displays the received parking status information. At this time, the parking lot management server 200 displays image information acquired in relation to the parking lot being managed by the parking lot management server 200 on one side of the screen so that the parking lot status can be easily checked (or grasped), and the screen It is also possible to display the received parking status information on the other side of the.

In addition, when the server 100 receives a destination transmitted from a terminal (not shown), a scheduled arrival date and time, identification information of a terminal, vehicle information, etc., the parking lot management server 200 is the server 100 And performing a parking reservation function for a specific parking lot having the highest possible parking availability searched in connection with the terminal.

In addition, the parking lot management server 200 receives encryption information transmitted from the server 100.

In addition, the parking lot management server 200 decrypts vehicle information, parking reservation function result information, and the like included in the received encrypted information using a private key managed by the corresponding parking lot management server 200.

In addition, when a vehicle reserved (or boarded) by a user of the corresponding terminal enters a parking lot managed by the parking management server 200 in the future, the parking management server 200 provides the vehicle information and the parking reservation function. Based on the result information, etc., it performs the entry/exit management function for the available parking space reserved by the corresponding terminal, the open/close management function of the blocker (not shown) at the entrance of the parking lot, and the simple parking fee settlement function. At this time, the parking lot management server 200 is based on payment information (eg, including Kakao Pay, Naver Pay, etc.) for simple payment included in the result information of the parking reservation function when a vehicle related to the corresponding terminal intends to leave the vehicle. It is possible to perform an automatic settlement function for parking charges non-face-to-face.

In addition, when the parking management server 200 sets the subscription function with the server 100, the parking management server 200 is performed on the server 100 through a dedicated app provided from the server 100 Various functions (e.g., pre-processing for image information, a function of applying pre-processed data to a pre-selected AI model to generate an output value, and a function of calculating information about parking spaces related to parking lots and performance measurement indicators of output values) , The function of determining whether or not parking is available for a parking space according to the calculated performance measurement index, and a function to reserve a parking space) can be independently performed.

In addition, the parking lot management server 200 may operate (or manage) a parking lot managed by the corresponding parking lot management server 200 through another dedicated app.

In this way, unstructured data for AI learning is collected, the collected unstructured data for AI learning is segmented, processed through the segmentation-processed unstructured data for AI learning, and trained a plurality of AI model architecture candidates, using performance measurement indicators. It is possible to measure the accuracy of multiple AI models, and select the optimal AI model based on the accuracy of the measured multiple AI models.

In addition, in this way, the image information acquired through the camera installed in the parking lot is processed, and the AI model selected before the processed image information is applied to recognize the parking space and the non-parking space within the image information. In addition, the available parking space and the non-parking space within the recognized image information can be displayed.

Hereinafter, a method of checking a parking space based on an image according to the present invention will be described in detail with reference to FIGS. 1 to 6.

3 to 4 are flowcharts illustrating a method of checking a parking space based on an image according to an exemplary embodiment of the present invention.

First, the server 100 receives image information transmitted (or provided) from the parking lot management server 200. Here, the image information is image information related to a parking lot managed by the parking lot management server 200, is obtained through a camera (not shown) installed at a fixed location in the parking lot, and is divided into a plurality of (or One or more) parking surfaces, and one or more cars parked on the parking surfaces. In this case, the server 100 may receive image information continuously transmitted from the parking lot management server 200 in a streaming format. Here, the parking management server 200 may be in a state in which a subscription function is set in relation to a parking space providing service provided by the server 100.

In addition, the server 100 performs preprocessing on the received image information (or streaming image information).

That is, the server 100 extracts a specific image (or a specific frame) from image information received in real time according to a preset analysis unit for a plurality of images (or a plurality of frames) constituting the received image information, and , Preprocessing is performed on the extracted specific image. Here, the pre-processing may be a function of converting pixels of the specific image (or image information) into an array, or a function of separating channels (eg, separating into RGB channels) for a plurality of pixels constituting the specific image. have.

In this way, the server 100 samples (or extracts) a specific image at a preset period (or analysis unit) among image information received in real time in order to reduce the load due to preprocessing, etc., and the sampled Pre-processing can be performed on a specific (or extracted) image.

For example, the server 100 receives first image information in a streaming format transmitted in real time from the first parking lot management server 200 that manages the ABCD parking lot.

In addition, the server 100 shows the first image information in the streaming form according to a preset analysis unit (for example, 1 frame every 2 seconds) for the received streaming form first image information. As illustrated, a first image 500 of 1920×1080 pixels is extracted, and 1920×1080 pixels of the extracted first image 500 are separated into RGB channels according to RGB values (S310).

Thereafter, the server 100 generates (or outputs) an output value according to AI modeling by applying the preprocessed data to a preselected (or set) AI model. Here, the AI model includes EfficientNet-L2, DeepLabv3+, and the like.

That is, the server 100 processes the preprocessed data as an input value of the AI model and applies it to the AI model, and then generates (or outputs) an output value according to the application of the AI model. Here, the output value is information indicating that a vehicle is included (for example, a preset '1') or information indicating that a vehicle is not included (for example, for each of the preprocessed data (or for each pixel in the preprocessed image)) For example, it includes a preset '0' or an inverse value of information indicating that the vehicle is included).

As an example, the server 100 applies the first image separated by the RGB channel as an input value of a pre-selected EfficientNet-L2 AI model, and a vehicle is included for each pixel as shown in FIG. 6. A first output value including information indicating that the vehicle is not included or information indicating that the vehicle is not included is generated (S320).

Thereafter, the server 100 calculates (or calculates) information on a plurality of parking spaces related to the corresponding parking lot stored in advance (or registered) in the server 100 and a performance measurement index of the output value according to the AI modeling. . At this time, the information on the parking space is location information on one or more parking spaces in relation to the parking lot corresponding to the parking lot management server 200, a labeling value (or unique identification information/number) corresponding to the location information, and parking It includes the total number of spaces (or total number of parking spaces/total number of parking spaces). In addition, the performance measurement indicators are IoU (Intersection over Union), precision, and reproducibility based on TP (True Positive), FP (False Positive), TN (True Negative), FN (False Negative), etc. Recall), accuracy, etc.

For example, the server 100 includes information on parking spaces related to the ABCD parking lot (for example, information on the first parking space to information on the 30th parking space) stored in advance in the server 100 and the EfficientNet -L2 Calculate performance measurement indexes (eg, first IoU to 30th IoU) between the first output values that are the output values of the AI model (S330).

Thereafter, the server 100 includes the calculated (or calculated) at least one performance measurement index (or the calculated information on the stored parking space and the performance measurement index of the output value according to the AI modeling) set in advance. Determine (or confirm) whether it is abnormal.

For example, the server 100 includes information on the calculated parking space related to the ABCD parking lot (for example, information on the first parking space to information on the 30th parking space) and the EfficientNet-L2 AI model. It is determined whether or not the first IoU to the 30th IoU between the first output values that are output values are equal to or greater than the first threshold value (S340).

When there is at least one performance measurement index that is equal to or greater than the threshold value among the determination result (or the confirmation result) and the calculated one or more performance measurement indexes, the server 100 may perform at least one performance measurement index related to the at least one performance measurement index. Determining one parking space (or at least one parking space corresponding to information on the parking space related to the at least one performance measurement index that is equal to or greater than the threshold value among the calculated one or more performance measurement indicators) as parking available (or Processing).

For example, the server 100 is the 4th IoU, 5th IoU, 6th IoU, 7th IoU, 11th IoU, 12th IoU, 13th IoU, of the calculated 1st IoU to 30th IoU. When the 14 IoU, the 17th IoU, the 20th IoU, the 24th IoU, the 28th IoU, the 29th IoU, and the 30th IoU are equal to or greater than the first threshold, the server 100 is the 4th IoU, the 5th IoU, Corresponding to the 6th IoU, 7th IoU, 11th IoU, 12th IoU, 13th IoU, 14th IoU, 17th IoU, 20th IoU, 24th IoU, 28th IoU, 29th IoU and 30th IoU Parking spaces 4 to 7, parking spaces 11 to 14, parking spaces 17, parking spaces 20, parking spaces 24, and parking spaces 28 to 30 can be parked It is determined as (S350).

In addition, the determination result (or the confirmation result), when there is at least one other performance measurement index less than the threshold value among the calculated one or more performance measurement indicators (or the threshold value among the calculated one or more performance measurement indicators) If there are other performance measurement indicators other than at least one performance measurement index that is the above), the server 100 measures at least one other parking space (or the calculated one or more performance measurements) related to the other at least one performance measurement index. Among the indicators, at least one other parking space corresponding to information on the parking space related to the other at least one performance measurement index that is less than the threshold value) is determined (or processed) as a state in which parking is impossible.

As an example, the server 100 includes the first IoU, the second IoU, the third IoU, the eighth IoU, the 9th IoU, the 10th IoU, the 15th IoU, among the calculated first IoU to 30th IoU, When the 16th IoU, 18th IoU, 19th IoU, 21st IoU, 22nd IoU, 23rd IoU, 25th IoU, 26th IoU, and 27th IoU are less than the first threshold, the server 100 is The 1st IoU, 2nd IoU, 3rd IoU, 8th IoU, 9th IoU, 10th IoU, 15th IoU, 16th IoU, 18th IoU, 19th IoU, 21st IoU, 22nd IoU, th 1st to 3rd parking spaces, 8th to 10th parking spaces, 15th to 16th parking spaces, 18th parking corresponding to 23 IoU, 25th IoU, 26th IoU and 27th IoU The space to the 19th parking space, the 21st to the 23rd parking space, and the 25th to 27th parking spaces are determined to be in a state in which parking is impossible (S360).

Thereafter, the server 100 according to the determination result, the parking status including parking spaces (or parking surfaces) and non-parking parking spaces among a plurality of parking surfaces related to the parking lot managed by the parking management server 200 Generate information. Here, the parking status information is the date and time information at the time of creation of the parking status information, the name of the parking lot, the location information of the parking lot, the contact information of the parking lot, the parking lot image (or the parking lot cross section/the above image information/image), the total number of parking spaces, It includes the number of parking spaces available for parking, the number of parking spaces that cannot be parked, location information (or labeling value/unique identification information/number) for parking spaces that can be parked, and location information for parking spaces that cannot be parked.

In addition, the server 100 transmits the generated parking status information to the parking management server 200.

For example, the server 100 may provide information on at least one parking space that is available for parking, information on at least one other parking space that cannot be parked, etc., determined for a plurality of parking spaces related to the ABCD parking lot. 1st date and time information (e.g. 08/01/2020 2:10 PM), 1st parking image, 1st parking lot name (e.g. ABCD parking lot), 1st Parking location information (e.g., 456 Seolleung-ro 123ga-gil, Gangnam-gu, Seoul), first contact information (e.g. 02-5678-1234), total number of parking spaces (e.g. 30), number of parking spaces ( For example, 14), the number of parking spaces that cannot be parked (for example, 16), location information about the parking spaces available for parking (for example, parking spaces 4 to 7, parking spaces 11 to 14 Space, the 17th parking space, the 20th parking space, the 24th parking space, and the 28th to 30th parking spaces), location information about the parking spaces that cannot be parked (for example, parking spaces 1 to 3, 8th to 10th parking spaces, 15th to 16th parking spaces, 18th to 19th parking spaces, 21st to 23rd parking spaces and 25th to 27th parking spaces) Generates first parking status information including, etc.

In addition, the server 100 transmits the generated first parking status information to the first parking management server (S370).

Thereafter, the parking lot management server 200 receives parking status information transmitted from the server 100.

In addition, the parking lot management server 200 displays the received parking status information. At this time, the parking lot management server 200 displays image information acquired in relation to the parking lot being managed by the parking lot management server 200 on one side of the screen so that the parking lot status can be easily checked (or grasped), and the screen It is also possible to display the received parking status information on the other side of the.

For example, the first parking lot management server receives first parking status information transmitted from the server 100.

In addition, the first parking lot management server displays streaming image information acquired in real time in relation to the ABCD parking lot on one side of the screen, and displays the received first parking status information on the other side of the screen (S380). ).

Thereafter, the server 100 collects vehicle flow data for each time slot in relation to a plurality of parking lots managed by the server 100. Here, the vehicle flow data for each time period includes the name of the parking lot, location information of the parking lot, the total number of parking spaces, and the number of parking spaces available for parking.

That is, the server 100 accumulates a plurality of parking status information for each parking lot that is generated from a plurality of image information related to a plurality of parking lots received in real time, and corresponds to the accumulated parking status information for each parking lot. Collects vehicle flow data for each parking lot (or vehicle flow data for each parking lot).

In addition, the server 100 calculates (or calculates) a parking availability probability for each of a plurality of parking lots based on the collected vehicle flow data for each time slot for the plurality of parking lots.

That is, the server 100 calculates the parking availability probability for each time slot, which is the probability of parking availability for each parking lot, based on the collected vehicle flow data for each of the plurality of parking lots.

For example, the server 100 accumulates a plurality of parking status information for each parking lot each generated from a plurality of image information related to the ABCD parking lot, AAAAA parking lot, and BBBBBB parking lot managed by the server 100, respectively, and the accumulated Based on the first parking status information or the 1000th parking status information related to the ABCD parking lot, as shown in Table 1 below, the probability of parking availability for each time slot related to the ABCD parking lot is calculated.

ABCD parking lot (%) AAAAA parking lot (%) BBBBBB parking lot (%) 00 o'clock to 01 o'clock 100 90 100 01 o'clock to 02 o'clock 100 100 100 02:00 ~ 03:00 100 100 100 03:00 ~ 04:00 100 100 100 04:00 ~ 5:00 100 100 100 05 o'clock to 06 o'clock 100 100 100 06:00 ~ 07:00 50 70 60 07:00 ~ 08:00 30 50 40 08:00 ~ 09:00 10 20 15 09:00 ~ 10:00 5 10 5 10 o'clock to 11 o'clock 5 10 5 11 o'clock to 12 o'clock 5 10 5 12 o'clock to 13 o'clock 5 10 5 13:00 ~ 14:00 10 15 5 14:00 ~ 15:00 20 15 15 15:00-16:00 40 30 20 16:00 ~ 17:00 50 30 25 17 o'clock to 18 o'clock 60 50 40 18:00-19:00 60 50 40 19:00-20:00 70 50 65 20:00-21:00 80 70 80 21:00-22:00 100 70 80 22:00-23:00 100 80 100 23 o'clock to 24 o'clock 100 90 100

In addition, the server 100 calculates a parking availability probability for each time slot related to the corresponding AAAAA parking lot, as shown in [Table 1], based on the 10001th parking status information or the 11000th parking status information related to the accumulated AAAAA parking lot. In addition, the server 100 calculates the parking availability probability for each time slot related to the BBBBBB parking lot, as shown in [Table 1], based on the accumulated parking status information of the 30001 to the 31000 parking status related to the accumulated BBBBBB parking lot. Do (S390).

Thereafter, when the server 100 receives a destination, a scheduled arrival date and time, and identification information of the terminal transmitted from the terminal (not shown), the server 100 Time when the terminal reaches the destination based on the probability of parking available for each parking lot, location information (or vehicle flow data by time), etc. of a plurality of parking lots, the received destination, and expected arrival date and time, etc. Time), search (or check) a specific parking lot (or a specific parking space within a specific parking lot) with the highest probability

At this time, the server 100 transmits the destination, the scheduled arrival date and time, the identification information of the terminal, vehicle information (eg, vehicle type, vehicle width, vehicle height, vehicle length, etc.) transmitted from the terminal. When receiving, the server 100 includes the probability of parking available for each time slot for a plurality of parking lots managed by the server 100, location information of the plurality of parking lots, the received destination, expected arrival date and time, and vehicle information. In consideration of together, it is also possible to search for a specific parking lot (or a specific parking space in a specific parking lot) with the highest possible parking probability at the time when the vehicle reserved (or on board) by the user with the corresponding terminal reaches the destination.

In addition, the server 100 performs a parking reservation function for a specific parking lot with the highest probability of the searched parking available in connection with the terminal and the parking management server 200 related to the searched specific parking lot.

At this time, when performing a parking reservation function for a specific parking lot with the highest probability of parking available, the server 100 generates encryption information by encrypting vehicle information and result information of the parking reservation function using a preset public key. , And transmits the generated encryption information to the parking lot management server 200 that has performed a reservation function. Here, the parking reservation function result information is date and time information at the time of reservation, parking reservation date and time information, reserved parking space information, reserved person information (eg, name, phone number, terminal identification information, etc.), simple payment It includes payment information (including, for example, Kakao Pay, Naver Pay, etc.).

In addition, the parking lot management server 200 receives encryption information transmitted from the server 100.

In addition, the parking lot management server 200 decrypts vehicle information, parking reservation function result information, and the like included in the received encrypted information using a private key managed by the corresponding parking lot management server 200.

In addition, when a vehicle reserved (or boarded) by a user of the corresponding terminal enters a parking lot managed by the parking management server 200 in the future, the parking management server 200 provides the vehicle information and the parking reservation function. Based on the result information, etc., it performs the entry/exit management function for the available parking space reserved by the corresponding terminal, the open/close management function of the blocker (not shown) at the entrance of the parking lot, and the simple parking fee settlement function. At this time, the parking lot management server 200 is based on payment information (eg, including Kakao Pay, Naver Pay, etc.) for simple payment included in the result information of the parking reservation function when a vehicle related to the corresponding terminal intends to leave the vehicle. It is possible to perform an automatic settlement function for parking charges non-face-to-face.

For example, the first destination transmitted from the server 100 from the first terminal (not shown), the first scheduled arrival date and time (for example, August 01, 2020 3:20 pm), the first vehicle When receiving information (eg, Hyundai Sonata Hybrid 8th generation, the first overall width 1860 cm, the first overall height 1445 cm, the first overall length 4900 cm, etc.), identification information of the first terminal, etc., the server 100 According to [Table 1], the probability of parking available for each time slot for a plurality of parking lots managed by the server 100, location information of the plurality of parking lots, etc., the received first destination, and the first expected date and time of arrival For example, August 01, 2020, 3:20 pm), based on the first vehicle information, the first terminal checks the ABCD parking lot with the highest possibility of parking at the time of reaching the first destination.

In addition, the server 100 performs a parking reservation function for the ABCD parking lot having the highest possible parking availability in connection with the first terminal and the first parking lot management server related to the searched ABCD parking lot. In this case, the server 100 may reserve a 30th parking space in which a corresponding vehicle can park according to the first vehicle information among a plurality of parking spaces included in the ABCD parking lot.

In addition, the server 100 generates first encrypted information by encrypting the first vehicle information, first parking reservation function result information, etc. using a first public key set in advance, and stores the generated first encryption information. It is transmitted to the first parking lot management server.

In addition, the first parking lot management server receives the first encryption information transmitted from the server 100.

In addition, the first parking lot management server decrypts the received first encryption information using a first private key corresponding to the first public key being managed by the corresponding first parking lot management server, Check the result information of the first parking reservation function.

In addition, when the 101st vehicle reserved by the user with the first terminal arrives at the entrance of the ABCD parking lot, the first parking lot management server is based on the first vehicle information, the first parking reservation function result information, etc. The 101st vehicle arriving at the entrance of the ABCD parking lot is checked, and the breaker at the entrance of the ABCD parking lot is maintained in an open state to induce the 101st vehicle to enter the ABCD parking lot.

In addition, after the 101st vehicle enters the ABCD parking lot, the 101st vehicle parks in a parking space available for parking (or the reserved 30th parking space).

In addition, when a vehicle is to be unloaded from the 101st vehicle, the first terminal interlocks with the first parking lot management server and a payment server (not shown) to provide a simple payment included in the first parking reservation function result information. 1 Performs a simple payment function for the first parking fee (for example, KRW 15,000) based on payment information (for example, Kakao Pay account information).

In addition, when the 101st vehicle is adjacent to the breaker after payment is normally performed, the first parking lot management server controls to maintain the breaker in an open state so that the 101st vehicle can exit the corresponding ABCD parking lot. (S400).

As described above, the embodiment of the present invention collects unstructured data for AI learning, performs segmentation processing on the collected unstructured data for AI learning, and selects a plurality of AI model architecture candidates through the segmentation-processed AI learning unstructured data. Train, measure the accuracy of multiple AI models using performance indicators, select the optimal AI model based on the accuracy of the measured multiple AI models, and accurately recognize the shape of the vehicle in the collected data can do.

In addition, as described above, an embodiment of the present invention processes image information acquired through a camera installed in a parking lot, and applies an AI model previously selected to the processed image information to Existing CCTV equipment without input of expensive equipment or manpower required to manage parking status in parking lot by recognizing available parking spaces and non-parking spaces, and displaying available parking spaces and non-parking spaces within the recognized image It is possible to grasp the current status of the entire parking lot and improve the efficiency of parking management operation by expanding it to a 24-hour unmanned system.

The above contents may be modified and modified without departing from the essential characteristics of the present invention by those of ordinary skill in the technical field to which the present invention pertains. Accordingly, the embodiments disclosed in the present invention are not intended to limit the technical idea of the present invention, but to explain the technical idea, and the scope of the technical idea of the present invention is not limited by these embodiments. The scope of protection of the present invention should be interpreted by the following claims, and all technical ideas within the scope equivalent thereto should be interpreted as being included in the scope of the present invention.

10: Video-based parking space identification system
100: server 200: parking management server

Claims (10)

A parking lot management server that acquires image information related to a parking lot under management and transmits the acquired image information; And
Perform pre-processing on the image information, apply the pre-processed data to a pre-selected AI model, generate an output value according to AI modeling, and a plurality of parkings related to the parking lot corresponding to the parking management server stored in advance in the server Calculate information about space and performance measurement index of the output value according to the AI modeling, determine whether the calculated one or more performance measurement indexes are equal to or greater than a preset threshold, and measure the calculated one or more performances as a result of the determination When there is at least one performance measurement index that is equal to or greater than the threshold value among the indicators, it is determined that at least one parking space related to the at least one performance measurement index can be parked, and as a result of the determination, the calculated at least one performance measurement When there is at least one other performance measurement index that is less than the threshold value among the indicators, it is determined that at least one other parking space related to the other at least one performance measurement index cannot be parked, and the parking lot management according to the determination result Including the server for generating parking status information including a parking space available and non-parking parking spaces among a plurality of parking surfaces related to a parking lot managed by the server, and transmitting the generated parking status information to the parking management server. Parking space identification system based on video. The method of claim 1,
The output value is,
Each of the preprocessed data includes information indicating that a vehicle is included or information indicating that a vehicle is not included. The method of claim 1,
The parking status information,
Date and time information at the time of creation of parking status information, parking lot name, parking lot location information, parking lot contact information, parking lot image, total number of parking spaces, number of parking spaces available, number of parking spaces that cannot be parked, parking spaces available for parking A system for identifying a parking space based on an image, comprising at least one of location information and location information on a parking space where parking is not possible. The method of claim 1,
The server,
In relation to the plurality of parking lots managed by the server, vehicle flow data for each time period are collected for each of the plurality of parking lots, and parking is possible for each of the plurality of parking lots based on the vehicle flow data for each time period for the collected parking lots. When calculating the probability and receiving the destination and the expected arrival date and time transmitted from the terminal, the probability of parking available by time slot for a plurality of parking lots managed by the server, location information of a plurality of parking lots, the received destination and the Based on the scheduled arrival date and time, the terminal searches for a specific parking lot with the highest probability of parking available at the time of reaching the destination, and interlocks with the terminal and a parking management server related to the searched specific parking lot, and the searched parking is possible. A parking space confirmation system based on an image, characterized in that a parking reservation function is performed for a specific parking lot with the highest probability. Performing, by the server, preprocessing on the image information transmitted from the parking lot management server;
Generating an output value according to AI modeling by applying the preprocessed data to a pre-selected AI model by the server;
Calculating, by the server, information on a plurality of parking spaces related to a parking lot corresponding to the parking lot management server stored in advance in the server and a performance measurement index of an output value according to the AI modeling;
Determining, by the server, whether the calculated at least one performance measurement index is equal to or greater than a preset threshold;
As a result of the determination, when there is at least one performance measurement index that is equal to or greater than the threshold value among the calculated one or more performance measurement indexes, the server may park at least one parking space related to the at least one performance measurement index. Determining the presence state;
As a result of the determination, when there is at least one other performance measurement index less than the threshold value among the calculated one or more performance measurement indicators, the server determines at least one other parking space related to the other at least one performance measurement index. Determining that parking is not possible;
Generating, by the server, parking status information including parking spaces that can be parked and parking spaces that cannot be parked from among a plurality of parking surfaces related to a parking lot managed by the parking management server according to the determination result; And
And transmitting, by the server, the generated parking status information to the parking management server. The method of claim 5,
The step of pre-processing the image information,
Extracting a specific image from image information received in real time according to a preset analysis unit for a plurality of images constituting the image information; And
And converting the pixels of the extracted specific image into an array or separating channels for a plurality of pixels constituting the specific image. The method of claim 5,
Collecting, by the server, vehicle flow data for each time slot for each of the plurality of parking lots in relation to the plurality of parking lots managed by the server;
Calculating, by the server, a parking availability probability for each time slot for each of the plurality of parking lots based on the collected vehicle flow data for each time slot for the plurality of parking lots;
When receiving the destination and the expected arrival date and time transmitted from the terminal, the server, the probability of parking available for each time slot for a plurality of parking lots managed by the server, location information of the plurality of parking lots, the received destination and the Searching for a specific parking lot having the highest parking availability probability at a time point at which the terminal reaches the destination based on a scheduled arrival date and time; And
By the server, interlocking with the terminal and the parking management server related to the searched specific parking lot, performing a parking reservation function for the specific parking lot with the highest probability of the searched parking available. How to check parking spaces based on the grounds. The method of claim 7,
The step of collecting vehicle flow data for each time slot for each of the plurality of parking lots,
Accumulating a plurality of parking status information for each parking lot generated from a plurality of image information related to a plurality of parking lots received in real time; And
And a process of collecting vehicle flow data for each time slot related to the parking lot based on the accumulated parking status information for each parking lot. The method of claim 7,
The vehicle flow data for each time period,
Parking space identification method based on an image comprising at least one of a parking lot name, location information of a parking lot, a total number of parking spaces, and a number of parking spaces available for parking. The method of claim 7,
Transmitting, by the server, the encrypted information generated by encrypting the vehicle information and the result information of the parking reservation function using a public key set in advance to the parking management server performing the reservation function;
Decrypting, by the parking lot management server, vehicle information and parking reservation function result information included in the encryption information using a private key managed by the parking lot management server; And
When the vehicle reserved by the user of the terminal wants to enter the parking lot managed by the parking lot management server, the parking lot reserved by the terminal is possible based on the vehicle information and the result information of the parking reservation function by the parking management server. A method for confirming a parking space based on an image, further comprising performing at least one of an entry/exit management function for a space, an open/close management function of a blocker at a parking lot entrance, and a simple parking fee settlement function.

Images