Video-based parking space navigation system

Abstract

The invention relates to the technical field of parking space navigation, in particular to a video-based parking space navigation system, which specifically comprises a video segmentation module, a parking space feature extraction module, an initial evaluation module, a sequence optimization module and a parking space candidate module, the problems that in the prior art, a strategy for accurately capturing and effectively recognizing the characteristics and states of the parking spaces is lacked, so that the parking space detection stability and channel characteristics cannot be comprehensively considered and scientific parking space recommendation and navigation cannot be given when an existing navigation system evaluates and sorts the unoccupied parking spaces of a large-scale parking lot are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of parking space navigation, and is a video-based parking space navigation system. Background Art

[0002] In the current urban transportation field, with the continuous increase in the vehicle ownership, the demand for intelligent management of parking lots has become increasingly prominent, and the video-based parking space navigation system has become an important technical means to solve the parking problem. However, there are still many significant problems in the existing such technologies. In terms of video processing, when the current system preprocesses the surveillance video, it is difficult to effectively remove the interference factors in the complex environment. Such as the common light changes, shadows and noise generated by the surveillance equipment itself in the parking lot will seriously interfere with the subsequent parking space recognition and analysis. Moreover, for the information such as watermarks and special symbols that may exist in the video, the existing technologies lack efficient removal methods, resulting in low quality of the processed video data. The accuracy of parking space recognition is also a major shortcoming of the existing technologies. The existing parking space recognition algorithms have poor adaptability to different types of parking lots. In some parking lots with complex layouts, dense parking spaces or special designs, the recognition accuracy drops significantly. At the same time, when the algorithm faces the situation that the parking space is partially blocked or the vehicle is parked irregularly, it often cannot accurately judge the state of the parking space, thus affecting the reliability of the entire navigation system. In the analysis of parking space information and the formulation of navigation strategies, the existing systems often only rely on a single factor for evaluation, such as the distance between the parking space and the entrance, while ignoring important information such as the channel conditions and the traffic flow around the parking space. This makes the recommended parking space may not be a truly convenient choice and cannot provide the most optimized parking solution for users. In addition, the existing navigation systems lack the ability of real-time dynamic adjustment and cannot update the navigation information in time according to the real-time changes of the parking spaces in the parking lot, resulting in situations such as the parking space has been occupied during the parking process of users, reducing the parking experience of users. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to address the problem in the prior art that there is a lack of strategies for accurately capturing and effectively identifying the characteristics and states of parking spaces, resulting in the inability to comprehensively consider the stability of parking space detection and channel characteristics during the evaluation and sorting of the navigation system, and unable to give scientific parking space recommendations and navigation. A video-based parking space navigation system is proposed.

[0004] In order to achieve the above object, the video-based parking space navigation system of the present invention includes the following modules:

[0005] Video segmentation module, parking space feature extraction module, initial evaluation module, sequence optimization module and parking space candidate module;

[0006] The video segmentation module is used to preprocess the real-time monitoring video of the parking lot, construct and train an AI vision model for parking space recognition, and divide the video into independent video segments according to different parking space partitions with the help of the AI vision model;

[0007] The parking space feature extraction module is used to set parking space status prompt words in the AI vision model, identify the parking space status information in the video, extract features such as the position and size of the parking space, and complete the extraction of parking space information;

[0008] The initial evaluation module is used to establish a plane coordinate system, count the number of effective vehicle appearance scenarios of each parking space in the video, evaluate the initial convenience value of each parking space, and mark the parking space information on the plane coordinate system according to this value;

[0009] The sequence optimization module is used to identify the channel information connecting each parking space area in the video, record the channel feature data, and determine the associated optimization coefficient between each parking space in the navigation alternative parking space sequence;

[0010] The parking space candidate module is used to synthesize the initial convenience value and the associated optimization coefficient, obtain the navigation suitability value of each parking space, sort the parking space information according to the navigation suitability value, and output the two-dimensional parking space candidate position map of the parking lot.

[0011] The video segmentation module is used to run the following specific steps:

[0012] S11: Preprocess the parking lot monitoring video, use a filtering algorithm to remove noise, remove special symbols and noise information such as HTML tags and watermarks through image recognition technology, and filter out interference elements such as irrelevant backgrounds and fixed facilities based on preset rules;

[0013] S12: Construct an AI vision model for parking space recognition based on a deep learning architecture, collect and label multi-scene parking lot monitoring video data as a training set. After completing the model training through parameter adjustment and algorithm optimization, divide the video into independent segments according to the parking partitions for subsequent parking space analysis.

[0014] The parking space information prompt words include: parking space status (idle, occupied), parking space position, parking space size, parking space number, parking space type (such as ordinary parking space, emergency parking space), and there are a total of R parking space partition information; the parking space feature extraction module is used to run the following specific steps:

[0015] S21: Map the parking space-related visual features in the parking lot monitoring video frame to a low-dimensional vector space through a visual feature embedding model, and train the visual feature embedding model through a visual feature learning algorithm to complete the self-learning between the parking space-related visual features to capture the association between different features; the parking space-related visual features include: the color, shape, and boundary of the parking space;

[0016] S22: Establish a visual feature analysis model, convert the independent parking lot surveillance video frame segments into a visual feature digital sequence X = (x1, x2... x i ... x I ) and input it into the visual feature analysis model to obtain the expression sequence of each visual feature in the video segment as Y = (y1, y2... y i ... y I ), where the total number of visual features in the video segment is I, and i is the subscript representing the i-th visual feature;

[0017] S23: According to the parking space information prompt words set for the visual feature analysis model, identify and calculate the feature vectors of each category of parking space information in the video. The calculation method of the feature vector of each category of parking space information is as follows: Among them, T r is the feature vector of the r-th category of parking space information, and d1, d2 ∈ {1, 2.. i};

[0018] S24: Extract the feature vectors of each category of parking space information to form a feature vector set of parking space information as ST = (T1, T2... T r );

[0019] S25: Input the feature vector set of parking space information into the convolutional layer and perform parking space feature extraction, and input the extracted features into the pooling layer to output the final entity representation of each category of parking space information. Among them, the size of the convolutional kernel is 10×10, and the height of each convolutional kernel is h.

[0020] The initial evaluation module is used to run the following steps:

[0021] S31: Take the vehicle entrance where the vehicle enters the parking lot as the center of the parking space information, and establish a coordinate system with this as the origin;

[0022] S32: At the same time, extract the appearance frequency data and appearance position data of each parking space in the surveillance video. The appearance frequency can reflect the stability of the detection of this parking space;

[0023] S33: Evaluate the initial convenience value of each parking space to the center of the parking space information. The evaluation strategy is specifically as follows: Among them, r, u represent the u-th parking space in the r-th partition, and r, v represent the v-th parking space information center in the r-th partition;

[0024] is the subscript, is the initial convenience value of the u-th parking space and the v-th parking space information center in the r-th partition; is the navigation distance between the u-th parking space and the v-th parking space information center in the r-th partition;

[0025] CS r,u ,CS r,v The number of valid vehicle appearance scenes with vehicles entering and exiting in the surveillance video around the u-th parking space in the r-th partition and around the v-th parking space information center;

[0026] S34: extracting the initial convenience value of each parking space and vehicle entrance, marking each parking space information into a plane coordinate system, arranging the initial convenience values ​​of all parking spaces in descending order, and intercepting the first 20% to form an initial navigation alternative parking space sequence.

[0027] The sequence optimization module includes: a channel feature extraction unit and a coefficient calculation unit;

[0028] The channel feature extraction unit is configured with the following strategy:

[0029] S41: Identify channel information connecting each parking area in the parking lot surveillance video, and synchronously record connection feature data of each channel; the channel connection features include: channel width features and traffic smoothness features, wherein the channel width features are obtained by analyzing the video image and converting the preset length unit, and the traffic smoothness features mainly consider whether there are obstacles and temporary parking of vehicles in the channel;

[0030] S42: extracting channel information with special traffic conditions in the video, and determining traffic properties, wherein the traffic properties include convenient traffic properties and obstructed traffic properties, wherein the convenient traffic properties are positive traffic conditions, such as wide channels, no obstacles, and small turning angles; and the obstructed traffic properties are negative traffic conditions, such as narrow channels, obstacles, or large turning angles.

[0031] The coefficient calculation unit is configured with the following strategy:

[0032] S43: Filter the information of two parking spaces closest to the passage in the parking lot surveillance video;

[0033] When the passage between two parking spaces is wide and unobstructed, based on the spatial position relationship of the parking spaces in the video image, with the driving direction of the vehicle entering the parking lot from the entrance as a reference, the parking space in front of the passage is the front parking space information, and the parking space behind the passage is the rear parking space information, and the first type of optimization strategy is implemented;

[0034] When the passage between two parking spaces is obstacle-free and the turning angle is small, the second optimization strategy is implemented based on the vehicle's driving direction, with the parking space in front of the passage as the front parking space information and the parking space behind the passage as the rear parking space information;

[0035] Conversely, when the passage between two parking spaces is narrow with obstacles or has a large turning angle, the parking space in front of the passage is the rear parking space information, and the parking space behind the passage is the front parking space information, and the third type of optimization strategy is executed;

[0036] S44: Calculate the k-th type of correlation optimization coefficient between two parking space information, and the specific calculation strategy is as follows:

[0037]

[0038] Among them, ΔL1 and ΔL2 are the navigation distances of the front parking space information and the rear parking space information in the initial navigation alternative parking space sequence from the center position of the passage respectively;

[0039] are the optimization factors of the k-th type of optimization strategy respectively,

[0040] The navigation suitability value of each parking space after completing the parking space correlation optimization is the product of the correlation optimization coefficient and the initial convenience value.

[0041] The thickness of the connecting lines between the parking spaces in the two-dimensional candidate position map of the parking lot matches the navigation suitability value from large to small, that is, the larger the navigation suitability value, the thicker the line connecting the parking spaces. Through this intuitive visualization method, users can more clearly understand the correlation between parking spaces and the suitability of each parking space.

[0042] Compared with the prior art, the technical effects of the present invention are as follows:

[0043] 1. The video segmentation module of the present invention comprehensively preprocesses the real-time monitoring video of the parking lot, removes interference elements such as noise, special symbols, and irrelevant backgrounds, and at the same time constructs and trains an AI vision model to segment the video by parking space area. This not only improves the quality of the video but also lays a foundation for accurately identifying parking spaces subsequently. The parking space feature extraction module can accurately capture features such as the color, shape, and boundary of the parking space by using an advanced visual feature embedding and analysis model, and combined with the set parking space information prompt words, can accurately identify information such as the state, position, and size of the parking space, greatly improving the accuracy of parking space recognition.

[0044] 2. The initial evaluation module of the present invention establishes a coordinate system with the vehicle entrance as the origin, comprehensively considers the appearance times and position data of the parking spaces in the video, as well as the navigation distance between the parking spaces and the entrance and the number of times of surrounding vehicle entry and exit scenarios, scientifically evaluates the initial convenience value of each parking space, and screens out the initial navigation alternative parking space sequence, providing users with a preliminary selection of high-quality parking spaces.

[0045] 3. The channel feature extraction unit of the sequence optimization module of the present invention identifies channel information and records features such as its width and traffic smoothness, and can also determine the traffic nature of the channel. The coefficient calculation unit executes different optimization strategies for adjacent parking spaces according to the channel conditions and calculates the associated optimization coefficients. This enables the navigation sequence to fully consider the actual channel conditions of the parking lot and better meet the actual parking needs.

[0046] 4. The parking space candidate module of the present invention combines the initial convenience value and the associated optimization coefficient to obtain the navigation suitability value, sorts the parking space information and outputs a two-dimensional parking space candidate position map, and the thickness of the connection line between parking spaces matches the navigation suitability value. This intuitive visualization method allows users to clearly understand the association and suitability between each parking space, greatly facilitating the users to quickly find a suitable parking space in the parking lot, improving the parking efficiency, and having good practical application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0048] Among them:

[0049] Figure 1 is a schematic structural diagram of the video-based parking space navigation system of the present invention;

[0050] Figure 2 is a schematic flow diagram of the operation steps of an initial evaluation module of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0051] In order to make the above objects, features and advantages of the present invention more obvious and understandable, the following will make a detailed description of the specific embodiments of the present invention in conjunction with the drawings of the specification.

[0052] In the following description, many specific details are set forth in order to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Those skilled in the art can make similar promotions without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0053] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure or characteristic that can be included in at least one implementation manner of the present invention. The appearances of "in one embodiment" in different places in this specification do not all refer to the same embodiment, nor are they separate or alternative embodiments that exclude each other with other embodiments.

[0054] Embodiment 1:

[0055] As Figure 1 , 2 shown, the video - based parking space navigation system according to the embodiment of the present invention, as Figure 1 shown, includes the following modules:

[0056] A video segmentation module, a parking space feature extraction module, an initial evaluation module, a sequence optimization module, and a parking space candidate module;

[0057] The video segmentation module is used to pre - process the real - time monitoring video of the parking lot, construct and train an AI vision model for parking space recognition, and divide the video into independent video segments according to different parking space partitions with the help of the AI vision model;

[0058] The video segmentation module is used to run the following specific steps:

[0059] S11: Pre - process the parking lot monitoring video, use a filtering algorithm to remove noise, remove special symbols such as HTML tags and watermarks and noise information through image recognition technology, and filter out interference elements such as irrelevant backgrounds and fixed facilities based on preset rules;

[0060] S12: Construct an AI vision model for parking space recognition based on a deep learning architecture, collect and label multi - scenario parking lot monitoring video data as a training set. After completing model training through parameter adjustment and algorithm optimization, divide the video into independent segments according to parking partitions for subsequent parking space analysis.

[0061] The parking space feature extraction module is used to set parking space status prompt words in the AI vision model, identify the parking space status information in the video, extract features such as the parking space position and size, and complete the extraction of parking space information;

[0062] The parking space information prompt words include: parking space status (idle, occupied), parking space position, parking space size, parking space number, parking space type (such as ordinary parking space, emergency parking space), and there are a total of R parking space partition information; The parking space feature extraction module is used to run the following specific steps:

[0063] S21: Map the parking - space - related visual features in the parking lot monitoring video frames to a low - dimensional vector space through a visual feature embedding model, and train the visual feature embedding model through a visual feature learning algorithm to complete the self - learning between parking - space - related visual features to capture the association between different features; The parking - space - related visual features include: the color, shape, and boundary of the parking space;

[0064] Exemplarily, in this embodiment, the visual feature learning algorithm is the Skip - gram algorithm.

[0065] S22: Establish a visual feature analysis model, convert independent parking lot surveillance video frame segments into a digital sequence of visual features X = (x1, x2... x i ... x I ) and input it into the visual feature analysis model to obtain the expression sequence of each visual feature in the video segment as Y = (y1, y2... y i ... y I ), where the total number of visual features in the video segment is I, and i is the subscript representing the i-th visual feature;

[0066] S23: According to the parking space information prompt words set for the visual feature analysis model, identify and calculate the feature vectors of each category of parking space information in the video. The calculation method of the feature vector of each category of parking space information is as follows: Among them, T r is the feature vector of the r-th category of parking space information, and d1, d2 ∈ {1, 2.. i};

[0067] S24: Extract the feature vectors of each category of parking space information to form a feature vector set of parking space information ST = (T1, T2... T r );

[0068] S25: Input the feature vector set of parking space information into the convolutional layer and perform parking space feature extraction, and input the extracted features into the pooling layer to output the final entity representation of each category of parking space information. Among them, the convolutional kernel size is 10×10, and the height of each convolutional kernel is h.

[0069] The initial evaluation module is used to establish a plane coordinate system, count the number of effective vehicle appearance scenarios of each parking space in the video, evaluate the initial convenience value of each parking space, and mark the parking space information on the plane coordinate system according to this value;

[0070] The initial evaluation module is used to perform the following steps:

[0071] S31: Use the vehicle entrance where the vehicle enters the parking lot as the center of parking space information, and establish a coordinate system with this as the origin;

[0072] S32: At the same time, extract the appearance times data and appearance position data of each parking space in the surveillance video. The appearance times can reflect the stability of the detection of this parking space;

[0073] S33: Evaluate the initial convenience value of each parking space and the center of parking space information. The evaluation strategy is as follows: Among them, r, u represent the u-th parking space in the r-th partition, and r, v represent the center of the v-th parking space information in the r-th partition;

[0074] is the subscript, is the initial convenience value of the u-th parking space and the v-th parking space information center in the r-th partition; is the navigation distance between the u-th parking space and the v-th parking space information center in the r-th partition;

[0075] CS r,u , CS r,v are respectively the number of effective vehicle appearance scenarios where vehicles enter and exit around the u-th parking space and the v-th parking space information center in the r-th partition in the surveillance video;

[0076] S34: Extract the initial convenience values of each parking space and the vehicle entrance, mark the information of each parking space in the plane coordinate system, sort the initial convenience values of all parking spaces in descending order, and intercept the first 20% to form the initial navigation alternative parking space sequence.

[0077] The sequence optimization module is used to identify the channel information connecting each parking space area in the video, record the channel feature data, and determine the correlation optimization coefficient between each parking space in the navigation alternative parking space sequence;

[0078] The sequence optimization module includes: a channel feature extraction unit and a coefficient calculation unit;

[0079] The channel feature extraction unit is configured with the following strategy:

[0080] S41: Identify the channel information connecting each parking space area in the parking lot surveillance video, and synchronously record the connection feature data of each channel; the connection features of the channel include: channel width feature and traffic smoothness feature, where the channel width feature is obtained by analyzing the video image in combination with a preset length unit conversion, and the traffic smoothness feature mainly considers whether there are obstacles and temporary vehicle stops in the channel;

[0081] S42: Extract the channel information with special traffic conditions in the video and perform traffic nature judgment. The traffic nature includes: convenient traffic nature and obstructive traffic nature. Among them, the convenient traffic nature is a positive traffic situation, such as a wide channel, no obstacles, and a small turning angle; the obstructive traffic nature is a negative traffic situation. Such as a narrow channel, obstacles, or a large turning angle, etc.

[0082] The coefficient calculation unit is configured with the following strategy:

[0083] S43: Screen the information of the two parking spaces closest to the channel in the parking lot surveillance video;

[0084] When the passage between two parking spaces is wide and unobstructed, based on the spatial position relationship of the parking spaces in the video image, with the driving direction of the vehicle entering the parking lot from the entrance as a reference, the parking space in front of the passage is the front parking space information, and the parking space behind the passage is the rear parking space information, and the first type of optimization strategy is implemented;

[0085] When the passage between two parking spaces is obstacle-free and the turning angle is small, the second optimization strategy is implemented based on the vehicle's driving direction, with the parking space in front of the passage as the front parking space information and the parking space behind the passage as the rear parking space information;

[0086] On the contrary, when the passage between two parking spaces is narrow and has obstacles or a large turning angle, the parking space in front of the passage is the rear parking space information, and the parking space behind the passage is the front parking space information, and the third type of optimization strategy is implemented;

[0087] S44: Calculate the k-th type of association optimization coefficient between two parking space information. The specific calculation strategy is as follows:

[0088]

[0089] Among them, ΔL1 and ΔL2 are the navigation distances between the front parking space information and the rear parking space information in the initial navigation candidate parking space sequence and the center position of the channel;

[0090] are the optimization factors of the k-th optimization strategy,

[0091] The parking space candidate module is used to combine the initial convenience value and the associated optimization coefficient to obtain the navigation suitability value of each parking space, sort the parking space information according to the navigation suitability value, and output a two-dimensional parking space candidate position map of the parking lot.

[0092] After the parking space association optimization is completed, the navigation suitability value of each parking space is the product of the association optimization coefficient and the initial convenience value.

[0093] Embodiment 2:

[0094] This embodiment provides an electronic device, including: a processor and a memory, wherein the memory stores a computer program that can be called by the processor;

[0095] The processor runs the above-mentioned video-based parking navigation system by calling the computer program stored in the memory.

[0096] The electronic device may vary greatly due to different configurations or performances, and can include one or more processors (Central Processing Units, CPUs) and one or more memories. Among them, at least one computer program is stored in the memory, and the computer program is loaded and run by the processor to implement the video-based parking space navigation system provided by the above method embodiments. The electronic device can also include other components for implementing device functions. For example, the electronic device can also have components such as wired or wireless network interfaces and input / output interfaces for data input and output. This embodiment will not be elaborated here.

[0097] Embodiment 3:

[0098] This embodiment provides a computer-readable storage medium, on which a rewritable computer program is stored;

[0099] When the computer program runs on a computer device, it causes the computer device to run the above video-based parking space navigation system.

[0100] For example, the computer-readable storage medium can be a read-only memory (ROM for short), a random access memory (RAM for short), a compact disc read-only memory (CD-ROM for short), magnetic tape, floppy disk, and optical data storage device, etc.

[0101] It should be understood that in various embodiments of the present application, the magnitudes of the serial numbers of the above processes do not mean the order of execution. The execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0102] It should be understood that determining B based on A does not mean determining B only based on A, and B can also be determined based on A and / or other information.

[0103] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. A computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions according to the embodiments of the present invention are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired network or / and a wireless network. The computer-readable storage medium can be any available medium that the computer can access, or a data storage device such as a server or data center that contains one or more collections of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0104] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed in the present invention can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0105] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be repeated here.

[0106] In several embodiments provided by the present invention, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of units is only one way, and there can be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of devices or units can be in electrical, mechanical, or other forms.

[0107] The unit described as a separation component may or may not be physically separated. The component displayed as a unit may or may not be a physical unit, that is, it may be located in one place or distributed to multiple network units. Some or all of these units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0108] In addition, each functional unit in various embodiments of the present invention can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit.

[0109] In the description of this specification, the descriptions referring to terms such as "one embodiment", "example", "specific example", etc. mean that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0110] In summary of the above embodiments, compared with the prior art, the technical effects of the present invention are as follows:

[0111] 1. The video segmentation module of the present invention comprehensively preprocesses the real-time monitoring video of the parking lot, removes interference elements such as noise, special symbols and irrelevant backgrounds, and at the same time constructs and trains an AI vision model to segment the video by parking space areas. This not only improves the quality of the video, but also lays a foundation for accurately identifying parking spaces subsequently. The parking space feature extraction module can accurately capture features such as the color, shape, and boundary of the parking space by using an advanced visual feature embedding and analysis model. Combining with the set prompt words of parking space information, it can accurately identify information such as the status, location, and size of the parking space, greatly improving the accuracy of parking space recognition.

[0112] 2. The initial evaluation module of the present invention establishes a coordinate system with the vehicle entrance as the origin, comprehensively considers the appearance times and position data of the parking spaces in the video, as well as the navigation distance between the parking spaces and the entrance and the number of times of surrounding vehicle entry and exit scenarios, scientifically evaluates the initial convenience value of each parking space, and screens out the initial navigation alternative parking space sequence, providing users with a preliminary selection of high-quality parking spaces.

[0113] 3. The channel feature extraction unit of the sequence optimization module of the present invention identifies channel information and records features such as its width and traffic smoothness, and can also judge the traffic nature of the channel. The coefficient calculation unit executes different optimization strategies on adjacent parking spaces according to the channel conditions and calculates the associated optimization coefficients. This enables the navigation sequence to fully consider the actual channel conditions of the parking lot and better meet the actual parking needs.

[0114] 4. The parking space candidate module of the present invention obtains the navigation suitability value by synthesizing the initial convenience value and the associated optimization coefficient, sorts the parking space information, and outputs a two-dimensional parking space candidate position map. Moreover, the thickness of the connection lines between parking spaces matches the navigation suitability value. This intuitive visualization method enables users to clearly understand the associations and suitability among various parking spaces, greatly facilitating the users to quickly find a suitable parking space in the parking lot, improving the parking efficiency, and having good practical application value.

[0115] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and the descriptions in the specification are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and all these changes and improvements fall within the scope of the present invention claimed. The scope of the present invention claimed is defined by the appended claims and their equivalents.

Claims

1. The video-based parking navigation system is characterized by: The system comprises: a video segmentation module, a parking space feature extraction module, an initial evaluation module, a sequence optimization module and a parking space candidate module; The video segmentation module is used to pre-process the real-time monitoring video of the parking lot, build and train an AI visual model for parking space recognition, and use the AI ​​visual model to segment the video into independent video segments according to different parking space partitions; The parking space feature extraction module is used to set parking space status prompt words in the AI ​​visual model, identify parking space status information in the video, extract parking space position and size, and complete parking space information extraction; The initial evaluation module is used to establish a plane coordinate system, count the number of valid vehicle appearance scenes of each parking space in the video, evaluate the initial convenience value of each parking space, and mark the parking space information into the plane coordinate system according to the value; The sequence optimization module is used to identify channel information connecting each parking space area in the video, record channel feature data, and determine the correlation optimization coefficient between each parking space in the navigation alternative parking space sequence; The parking space candidate module is used to combine the initial convenience value and the associated optimization coefficient to obtain the navigation suitability value of each parking space, sort the parking space information according to the navigation suitability value, and output a two-dimensional parking space candidate position map of the parking lot.

2. The video-based parking navigation system according to claim 1, characterized in that: The video segmentation module is used to run the following specific steps: S11: Pre-process the parking lot surveillance video, use filtering algorithms to remove noise, use image recognition technology to remove special symbols and noise information, and filter irrelevant background and interference elements based on preset rules; S12: Build an AI vision model for parking space recognition based on deep learning architecture, and collect and annotate multi-scene parking lot surveillance video data as a training set; After completing the model training through parameter adjustment and algorithm optimization, the video is divided into independent segments according to the parking partitions.

3. The video-based parking navigation system according to claim 2, characterized in that: The parking space feature extraction module is used to run the following specific steps: S21: Map the parking space related visual features in the parking lot monitoring video frame to a low-dimensional vector space through a visual feature embedding model, and train the visual feature embedding model through a visual feature learning algorithm to complete self-learning between parking space related visual features; S22: Establish a visual feature analysis model to convert independent parking lot surveillance video frame segments into visual feature digital sequences X = (x1, x2...x i ...x I ) and input it into the visual feature analysis model to obtain the expression sequence of each visual feature in the video clip as Y = (y1, y2...y i ...y I ), where the total number of visual features in the video clip is I, and i is a subscript, indicating the i-th visual feature; S23: According to the parking space information prompt words set for the visual feature analysis model, the feature vectors of each category of parking space information in the video are identified and calculated. The feature vector of each category of parking space information is calculated as follows: Among them, T r is the feature vector of the r-th type of parking space information, d1,d2∈{1,2..i}; S24: Extract the feature vector of each category of parking space information to form a feature vector set of parking space information: ST = (T1, T2...T r ); S25: Input the feature vector set of the parking space information into the convolution layer and extract the parking space features, and input the extracted features into the pooling layer to output the final entity representation of various types of parking space information, wherein the convolution kernel size is 10×10, and the height of each convolution kernel is h.

4. The video-based parking navigation system according to claim 3, characterized in that: The initial assessment module is used to run the following steps: S31: Taking the vehicle entrance into the parking lot as the parking space information center, and establishing a coordinate system with this as the origin; S32: extracting the appearance number data and the appearance location data of each parking space in the surveillance video at the same time; S33: Evaluate the initial convenience value of each parking space and the parking space information center. The evaluation strategy is as follows: Among them, r,u represents the u-th parking space in the r-th partition, and r,v represents the v-th parking space information center in the r-th partition; is the subscript, is the initial convenience value of the u-th parking space and the v-th parking space information center in the r-th partition; is the navigation distance between the u-th parking space and the v-th parking space information center in the r-th partition; CS r,u ,CS r,v The number of valid vehicle appearance scenes with vehicles entering and exiting in the surveillance video around the u-th parking space in the r-th partition and around the v-th parking space information center; S34: extracting the initial convenience value of each parking space and vehicle entrance, marking each parking space information into a plane coordinate system, arranging the initial convenience values ​​of all parking spaces in descending order, and intercepting the first 20% to form an initial navigation alternative parking space sequence.

5. The video-based parking navigation system according to claim 4, characterized in that: The sequence optimization module includes: a channel feature extraction unit and a coefficient calculation unit; The channel feature extraction unit is configured with the following strategy: S41: Identify channel information connecting each parking area in the parking lot surveillance video, and synchronously record connection feature data of each channel; the channel connection features include: channel width features and traffic smoothness features, wherein the channel width features are obtained by analyzing the video image and converting the preset length unit, and the traffic smoothness features mainly consider whether there are obstacles and temporary parking of vehicles in the channel; S42: extracting channel information of special traffic conditions in the video, and determining the traffic nature, wherein the traffic nature includes: convenient traffic nature and obstructed traffic nature, wherein the convenient traffic nature is a positive traffic condition, and the obstructed traffic nature is a negative traffic condition.

6. The video-based parking navigation system according to claim 5, characterized in that: The coefficient calculation unit is configured with the following strategy: S43: Filter the information of two parking spaces closest to the passage in the parking lot surveillance video; When the passage between two parking spaces is wide and unobstructed, based on the spatial position relationship of the parking spaces in the video image, with the driving direction of the vehicle entering the parking lot from the entrance as a reference, the parking space in front of the passage is the front parking space information, and the parking space behind the passage is the rear parking space information, and the first type of optimization strategy is implemented; When the passage between two parking spaces is obstacle-free and the turning angle is small, the second optimization strategy is implemented based on the vehicle's driving direction, with the parking space in front of the passage as the front parking space information and the parking space behind the passage as the rear parking space information; On the contrary, when the passage between two parking spaces is narrow and has obstacles or a large turning angle, the parking space in front of the passage is the rear parking space information, and the parking space behind the passage is the front parking space information, and the third type of optimization strategy is implemented; S44: Calculate the k-th type of association optimization coefficient between two parking space information. The specific calculation strategy is as follows: Among them, ΔL1 and ΔL2 are the navigation distances between the front parking space information and the rear parking space information in the initial navigation candidate parking space sequence and the center position of the channel; are the optimization factors of the k-th optimization strategy, 7. The video-based parking navigation system according to claim 6, characterized in that: After the parking space association optimization is completed, the navigation suitability value of each parking space is the product of the association optimization coefficient and the initial convenience value.

Images