New energy parking lot intelligent management system and method

Abstract

The invention relates to the field of intelligent management of parking lots, and provides an intelligent management system and method for a new energy parking lot. Whether the target analysis driving-in vehicle is a fuel vehicle or not is judged by performing feature analysis of a high-dimensional space on a license plate area in the target analysis driving-in vehicle image, and if the target analysis driving-in vehicle is the fuel vehicle, voice broadcast prompt is performed and parking lock lifting operation is executed. Therefore, the accuracy of vehicle type identification can be remarkably improved, and the maintenance of the specificity of the new energy parking spaces in the parking lot is facilitated.

Description

Technical Field

[0001] The present application relates to the field of intelligent parking lot management, and more specifically, to a new energy parking lot intelligent management system and method. Background Art

[0002] New energy parking spaces are usually equipped with charging facilities such as charging piles. The occupation of fuel vehicles will hinder the timely charging of new energy vehicles and affect the travel plans of new energy vehicle owners. Especially for owners who are running out of power, it may cause great inconvenience and even "range anxiety". In addition, new energy parking spaces are planned and set up to meet the special charging needs of new energy vehicles. The long-term occupation of fuel vehicles violates the principle of special resources, causes waste of resources, and reduces the overall service efficiency of the parking lot. Therefore, many parking lots now have implemented a ban on fuel vehicles entering new energy parking spaces. In actual parking lot scenes, light conditions are complex and changeable. Traditional license plate recognition algorithms are prone to inaccurate license plate recognition under different lighting environments (such as direct strong light and shadow parts). This leads to the possibility of misidentifying new energy vehicles as fuel vehicles or vice versa, which will bring inconvenience to new energy vehicle owners and cause confusion in the allocation of parking space resources in parking lots.

[0003] Therefore, a new energy parking lot intelligent management solution is needed. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present application provides a new energy parking lot intelligent management system and method.

[0005] A new energy parking lot intelligent management method, comprising:

[0006] Acquire an image of a vehicle entering the vehicle for target analysis captured by a camera;

[0007] Performing preliminary preprocessing on the image of the target analysis vehicle entering the vehicle to obtain a target analysis vehicle license plate interest region;

[0008] Analyze the target analysis vehicle license plate region of interest to obtain a target analysis vehicle license plate color fusion feature map;

[0009] According to the target, the information in the vehicle license plate color fusion feature map is analyzed to determine whether a voice broadcast prompt and a parking lock raising operation are required.

[0010] A new energy parking lot intelligent management system, comprising:

[0011] A vehicle image data acquisition module, used to acquire an image of a vehicle entering the vehicle for target analysis captured by a camera;

[0012] A license plate region extraction module is used to pre-process the image of the target analyzed vehicle to obtain a license plate region of interest of the target analyzed vehicle;

[0013] A license plate color feature analysis module is used to analyze the target analysis vehicle license plate region of interest to obtain a target analysis vehicle license plate color fusion feature map;

[0014] The vehicle type response module is used to analyze the information in the vehicle license plate color fusion feature map according to the target, and determine whether it is necessary to perform a voice broadcast prompt and execute the parking lock raising operation.

[0015] This application has significant technical effects due to the adoption of the above technical solutions:

[0016] The intelligent management system and method for new energy parking lots provided in this application adopts computer vision-based image processing technology, and performs high-dimensional spatial feature analysis on the license plate area in the image of the target vehicle to determine whether the target vehicle is a fuel vehicle. If it is a fuel vehicle, a voice broadcast prompt is given and the parking lock is raised. In this way, the accuracy of vehicle type recognition can be significantly improved, which helps to maintain the exclusivity of new energy parking spaces in parking lots. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0018] Figure 1 Flow chart of the intelligent management method of a new energy parking lot according to an embodiment of the present application.

[0019] Figure 2 This is a flowchart of step S2 in the new energy parking lot intelligent management method according to an embodiment of the present application.

[0020] Figure 3 This is a flowchart of step S3 in the new energy parking lot intelligent management method according to an embodiment of the present application.

[0021] Figure 4 This is a flowchart of step S4 in the new energy parking lot intelligent management method according to an embodiment of the present application.

[0022] Figure 5 This is a system block diagram of a new energy parking lot intelligent management system according to an embodiment of the present application. DETAILED DESCRIPTION

[0023] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0024] With the global emphasis on environmental protection and sustainable development, the promotion and application of new energy vehicles has gradually become an important direction for the development of urban transportation. In order to support the development of new energy vehicles and meet their special charging needs, many parking lots have specially set up new energy parking spaces, which are usually equipped with necessary charging facilities such as charging piles. However, in actual operation, the problem of fuel vehicles occupying parking spaces often occurs, which not only prevents new energy vehicles from obtaining the required power replenishment in a timely manner, but also may disrupt the travel plans of car owners, especially when the vehicle is about to run out of power, which may lead to "range anxiety", that is, the anxiety caused by the worry that the vehicle will not be able to find a charging station before the power runs out.

[0025] From the perspective of resource management, new energy parking spaces are specially planned and set up to serve new energy vehicles. Their existence is to ensure that such vehicles can be charged conveniently, thereby promoting the development of green travel. If fuel vehicles occupy these parking spaces for a long time, it will violate the principle of special resources for special purposes, cause unnecessary waste of resources, and also reduce the overall service efficiency and service quality of the parking lot.

[0026] To solve this problem, many parking lots have taken measures to prohibit fuel vehicles from entering new energy parking spaces. However, in the actual implementation process, new challenges are faced. In real scenes, the light conditions are complex and changeable (such as strong direct light and shadows), which poses a great challenge to the traditional license plate recognition algorithm that parking lots rely on. In this case, it is easy to cause misjudgment, such as misidentifying new energy vehicles as fuel vehicles or vice versa, which brings inconvenience to new energy vehicle owners and may also cause confusion in the allocation of parking spaces in the parking lot. Therefore, a smart management solution for new energy parking lots is expected.

[0027] In recent years, deep learning and neural networks have been widely used in computer vision, natural language processing, text signal processing and other fields. In addition, deep learning and neural networks have also shown a level close to or even beyond that of humans in image classification, object detection, semantic segmentation, text translation and other fields. The development of deep learning and neural networks provides new solutions and solutions for the intelligent management of new energy parking lots.

[0028] Figure 1 FIG. 1 is a flow chart of a new energy parking lot intelligent management method according to an embodiment of the present application. Figure 1As shown, according to the embodiment of the present application, the intelligent management method of the new energy parking lot includes: S1, acquiring an image of the target analysis vehicle entering by a camera; S2, performing preliminary preprocessing on the image of the target analysis vehicle entering to obtain a target analysis vehicle license plate interest region; S3, analyzing the target analysis vehicle license plate interest region to obtain a target analysis vehicle license plate color fusion feature map; S4, judging whether it is necessary to perform a voice broadcast prompt and execute a parking lock lifting operation according to the information in the target analysis vehicle license plate color fusion feature map.

[0029] In step S1, an image of a target analysis vehicle entering the vehicle captured by a camera is obtained. It should be understood that the captured image of the target analysis vehicle entering the vehicle contains vehicle appearance information and license plate information. The license plate information is the most critical part, including the base color of the license plate and the color and font of the license plate characters. The license plate rules in different regions are different, but generally speaking, there are certain differences between the license plates of new energy vehicles and fuel vehicles. By accurately detecting and color analyzing the license plate area in the image, the type of vehicle can be determined.

[0030] In step S2, the image of the vehicle entering the target analysis is pre-processed to obtain the target analysis vehicle license plate interest region. Specifically, Figure 2 FIG. 1 is a flow chart of step S2 in the new energy parking lot intelligent management method according to an embodiment of the present application. Figure 2 As shown, the step S2 includes: S21, performing illumination compensation on the image of the target analyzed vehicle to obtain the target analyzed vehicle image after illumination compensation; S22, passing the target analyzed vehicle image after illumination compensation through a license plate target detection network to obtain the target analyzed vehicle license plate region of interest.

[0031] In step S21, the image of the target analyzed vehicle entering the vehicle is subjected to illumination compensation to obtain the target analyzed vehicle image after illumination compensation. Accordingly, considering that in the actual parking environment, the illumination conditions are complex and changeable. The vehicle image may be affected by natural light (such as sunlight at different angles, including illumination under different weather conditions such as early morning, noon, evening, cloudy days, etc.) and artificial lighting (such as the lighting layout, light intensity and color at different locations in the parking lot). These illumination conditions will cause the image to have too bright highlight areas and too dark shadow areas, making the image brightness uneven. For the image of the target analyzed vehicle entering the vehicle, this uneven illumination will reduce the image quality and affect the subsequent extraction and recognition of key information such as the license plate. For example, the color of the license plate, the clarity of the characters, and the recognition of the edge of the license plate may be seriously distorted under uneven illumination. In order to reduce the image quality problems caused by uneven illumination and make the target object (mainly the license plate) in the image more uniform in brightness and contrast, it is necessary to perform illumination compensation on the image of the target analyzed vehicle entering the vehicle in this application. After light compensation, the color characteristics of the license plate area will be more stable and clear. In other words, for distinguishing new energy vehicles from fuel vehicles, light compensation can restore the true color of the license plate and avoid color misjudgment caused by light.

[0032] In step S22, the illumination-compensated target analysis vehicle image is passed through a license plate target detection network to obtain the target analysis vehicle license plate interest region. Accordingly, considering that in the complex environment of a parking lot, the overall image of the vehicle contains a large amount of information, such as the body, wheels and other parts, and the most critical part of judging the vehicle type is the vehicle license plate information, in order to exclude some irrelevant information and make the subsequent analysis more focused on the information directly related to the vehicle type judgment, in this application, the illumination-compensated target analysis vehicle image needs to be input into the license plate target detection network for processing. The license plate target detection network is constructed based on a deep learning algorithm. After being trained with a large amount of training data, it can quickly and accurately find the license plate area in the image and extract it for processing. Compared with the traditional method based on manual features (such as using edge detection), it has higher accuracy and robustness, and can adapt to complex situations such as different illumination, different license plate styles and different vehicle postures. It is also worth mentioning that processing the entire vehicle image requires a lot of computing resources, and in practical applications, computing resources are often limited. By processing only the area of ​​interest of the license plate, computing resources can be concentrated in key areas, avoiding wasting resources in non-critical areas (such as vehicle body, wheels, etc.), thereby improving the computational efficiency of the model and enabling the task of determining the vehicle type to be completed more quickly with limited hardware resources.

[0033] In step S3, the target vehicle license plate region of interest is analyzed to obtain a target vehicle license plate color fusion feature map. Specifically, Figure 3 FIG. 1 is a flow chart of step S3 in the new energy parking lot intelligent management method according to an embodiment of the present application. Figure 3 As shown, the step S3 includes: S31, extracting the license plate color features of the target analysis vehicle license plate area of ​​interest to obtain a target analysis vehicle license plate color feature map; S32, emphasizing the license plate color features of the target analysis vehicle license plate color feature map to obtain a target analysis vehicle license plate color feature enhancement map; S33, fusing the target analysis vehicle license plate color feature map and the target analysis vehicle license plate color feature enhancement map to obtain the target analysis vehicle license plate color fusion feature map.

[0034] In step S31, the license plate color feature extraction is performed on the target analysis vehicle license plate region of interest to obtain the target analysis vehicle license plate color feature map. Specifically, in the embodiment of the present application, the step S31 includes: passing the target analysis vehicle license plate region of interest through a license plate color feature encoder to obtain the target analysis vehicle license plate color feature map. It should be understood that the license plate color is an important discriminant factor when distinguishing between new energy vehicles and fuel vehicles. In order to extract the license plate color feature from the target analysis vehicle license plate region of interest, the target analysis vehicle license plate region of interest needs to be input into the license plate color feature encoder for processing. Here, the license plate color feature encoder is a convolutional neural network model as a feature extractor. Those of ordinary skill in the art should know that a convolutional neural network (CNN) has the ability to automatically learn and extract image features. For the license plate color feature encoder, it can automatically learn the pattern and law of color distribution from the license plate region of interest without manually designing complex color feature extraction rules. Moreover, in the actual parking environment, the license plate image may be disturbed by various factors, such as uneven lighting, scratches or stains on the license plate, blurred images, etc. The convolutional neural network has learned the ability to deal with these complex situations through a large amount of training data. It can ignore these interference factors to a certain extent and focus on extracting essential features related to color. For example, even if part of the license plate is blocked by shadows, the model can still infer the color characteristics of the entire license plate through the color information of the unblocked part and the associated information of the surrounding area.

[0035] In step S32, the license plate color feature of the target analysis vehicle license plate color feature map is emphasized to obtain the target analysis vehicle license plate color feature enhancement map. Specifically, in the embodiment of the present application, the step S32 includes: passing the target analysis vehicle license plate color feature map through a license plate color feature enhancer to obtain the target analysis vehicle license plate color feature enhancement map. It should be understood that in the actual image acquisition and processing process, although the license plate color feature map has been extracted by the color feature encoder, due to various interference factors, such as image noise, partial color information fuzziness, etc., the license plate color feature may not be prominent enough. In order to enhance the significance of key color features so that these features are easier to be identified and used in subsequent vehicle type judgment, it is necessary to emphasize the license plate color feature of the target analysis vehicle license plate color feature map. Specifically, the target analysis vehicle license plate color feature map is input into the license plate color feature enhancer. By emphasizing the color features, the difference between the colors of different types of license plates can be magnified, thereby improving the ability to distinguish vehicle types. Here, the license plate color feature enhancer is a convolutional neural network model using a channel attention mechanism. The channel attention mechanism can adaptively learn the importance of each channel (in the color feature map, the channel can be understood as information of different color dimensions). This means that the model can automatically determine which color channels are more critical for distinguishing vehicle types based on the specific content of the input license plate color feature map, and perform targeted enhancements on these channels. For example, if the green channel is determined to be the most important channel in the color features of new energy license plates, the model will automatically enhance the information of the green channel instead of enhancing all color channels to the same extent, thereby achieving more accurate feature emphasis.

[0036] In step S33, the target analysis vehicle license plate color feature map and the target analysis vehicle license plate color feature enhancement map are fused to obtain the target analysis vehicle license plate color fusion feature map. It should be understood that the target analysis vehicle license plate color feature map contains the original color feature information extracted from the license plate area. This information is obtained based on the color distribution of the license plate itself and has basic representativeness; while the target analysis vehicle license plate color feature enhancement map is the result of emphasizing the key color features through the color feature enhancer on the basis of the original color features; that is, the original color feature map provides comprehensive color distribution details, while the license plate color feature enhancement map highlights the more critical color feature part in vehicle type judgment. In order to combine these two types of information so that the final feature map contains complete color information and highlights important distinguishing features, it is necessary to fuse the target analysis vehicle license plate color feature map and the target analysis vehicle license plate color feature enhancement map in this application. The fused feature map can provide richer and more accurate color information, which helps to more accurately judge the vehicle type.

[0037] In step S4, the information in the vehicle license plate color fusion feature map is analyzed according to the target to determine whether a voice broadcast prompt and a parking lock raising operation need to be performed. Specifically, Figure 4 FIG. 4 is a flow chart of step S4 in the new energy parking lot intelligent management method according to an embodiment of the present application. Figure 4 As shown, the step S4 includes: S41, performing semantic area focusing optimization based on high-dimensional feature deconstruction on the target analysis vehicle license plate color fusion feature map to obtain an optimized target analysis vehicle license plate color fusion feature map; S42, passing the optimized target analysis vehicle license plate color fusion feature map through a vehicle category classifier to obtain a vehicle classification result, and the vehicle classification result is used to indicate whether the target analysis entering vehicle is a new energy vehicle or a fuel vehicle; S43, in response to the vehicle classification result that the target analysis entering vehicle is a fuel vehicle, performing a voice broadcast prompt and executing a parking lock lifting operation.

[0038] In step S41, the target analysis vehicle license plate color fusion feature map is optimized by semantic region focusing based on high-dimensional feature deconstruction to obtain an optimized target analysis vehicle license plate color fusion feature map. In particular, considering that there may be repeated or low-correlation features between different channels of the target analysis vehicle license plate color fusion feature map, these redundant information will not increase the effectiveness of classification, but may introduce noise and affect the generalization ability of the model. For the task of judging whether a vehicle is a new energy vehicle or a fuel vehicle, too many redundant features may make it difficult for the model to focus on those key features that are truly helpful for classification, thereby reducing the classification accuracy. In order to improve the accuracy of vehicle type recognition, it is necessary to optimize the target analysis vehicle license plate color fusion feature map to highlight those key features for vehicle type recognition. However, traditional deep learning feature optimization methods usually focus on adjusting local pixel weights, but ignore the relationship between the overall spatial context distribution in the feature map, that is, they cannot well understand the overall distribution and mutual relationship of license plate colors, and they may miss some important global features, such as the systematic differences in design between new energy license plates and fuel license plates, which may be the key to distinguishing between the two types of vehicles. Based on this, in the technical solution of the present application, it is necessary to perform semantic area focusing optimization based on high-dimensional feature deconstruction on the target analysis vehicle license plate color fusion feature map to obtain an optimized target analysis vehicle license plate color fusion feature map.

[0039] Specifically, in an embodiment of the present application, the step S41 includes: performing feature deconstruction along the channel dimension on the target analysis vehicle license plate color fusion feature map to obtain a set of target analysis vehicle license plate color fusion feature vectors; performing feature decomposition based on eigenvalues ​​on each target analysis vehicle license plate color fusion feature vector in the set of target analysis vehicle license plate color fusion feature vectors to obtain a set of target analysis vehicle license plate color fusion principal component intrinsic coding vectors; calculating the spatial distance entropy between each group of corresponding target analysis vehicle license plate color fusion feature vectors and target analysis vehicle license plate color fusion principal component intrinsic coding vectors in the set of target analysis vehicle license plate color fusion feature vectors and the set of target analysis vehicle license plate color fusion principal component intrinsic coding vectors to obtain a target analysis vehicle license plate color recognition spatial constraint vector composed of multiple spatial distance entropies; calculating the product between the target analysis vehicle license plate color recognition spatial constraint vector and its transposed vector to obtain a target analysis vehicle license plate color recognition spatial constraint matrix; inputting the target analysis vehicle license plate color recognition spatial constraint matrix into the S based The spatial constraint activation unit of the igmoid activation function is used to obtain the target analysis vehicle license plate color recognition spatial constraint feature matrix; based on the target analysis vehicle license plate color recognition spatial constraint feature matrix, the target analysis vehicle license plate color fusion feature map is feature constrained to obtain the optimized target analysis vehicle license plate color fusion feature map.

[0040] More specifically, in an embodiment of the present application, each target analysis vehicle license plate color fusion feature vector in the set of the target analysis vehicle license plate color fusion feature vector is subjected to eigenvalue-based eigendecomposition to obtain a set of target analysis vehicle license plate color fusion principal component eigencoding vectors, including: processing the set of the target analysis vehicle license plate color fusion feature vectors according to the following formula to obtain the set of the target analysis vehicle license plate color fusion principal component eigencoding vectors; wherein the formula is:

[0041]

[0042] Among them, V i represents the i-th target analysis vehicle license plate color fusion feature vector in the set of the target analysis vehicle license plate color fusion feature vectors, PCA (V i ) indicates the value of V i Perform eigenvalue-based eigendecomposition, U i Yes V i The corresponding target analysis vehicle license plate color fusion eigendecomposition vector sequence, Λ i V i The corresponding target analysis vehicle license plate color fusion diagonal matrix, U i T For Ui The transpose of v i1 、v i2 、v im V i The first, second and mth target analysis vehicle license plate color fusion intrinsic decomposition vectors of the corresponding target analysis vehicle license plate color fusion intrinsic decomposition vectors, λ i1 , im V i The corresponding target analysis vehicle license plate color fusion diagonal matrix of the first and m-th position eigenvalues, V i ' indicates V i The corresponding target analysis vehicle license plate color is fused with the principal component eigenvalue encoding vector.

[0043] More specifically, in an embodiment of the present application, calculating the spatial distance entropy between each corresponding target analysis vehicle license plate color fusion feature vector and the target analysis vehicle license plate color fusion principal component intrinsic coding vector in the set of the target analysis vehicle license plate color fusion feature vectors and the set of the target analysis vehicle license plate color fusion principal component intrinsic coding vectors to obtain a target analysis vehicle license plate color recognition spatial constraint vector composed of multiple spatial distance entropies, including: processing the set of the target analysis vehicle license plate color fusion feature vectors and the set of the target analysis vehicle license plate color fusion principal component intrinsic coding vectors according to the following formula to obtain the target analysis vehicle license plate color recognition spatial constraint vector; wherein, the formula is:

[0044]

[0045] Among them, V i represents the i-th target analysis vehicle license plate color fusion feature vector in the set of target analysis vehicle license plate color fusion feature vectors, V i ' indicates V i The corresponding target analysis vehicle license plate color fusion principal component eigencode vector, ||·||2 2 is the square of the Euclidean norm of the vector, arccosh is the inverse hyperbolic cosine function, d P (V i ,V i ') indicates V i and V i 'The spatial distance entropy between i Represents the i-th spatial distance entropy in the target analysis vehicle license plate color recognition space constraint vector.

[0046] In view of the above technical problems, in the technical solution of the present application, the semantic region focusing optimization based on high-dimensional feature deconstruction is performed on the target analysis vehicle license plate color fusion feature map. The process begins with the feature deconstruction of the target analysis vehicle license plate color fusion feature map along the channel dimension to obtain a set of target analysis vehicle license plate color fusion feature vectors. It should be understood that by deconstructing the target analysis vehicle license plate color fusion feature map, the features of different channels can be extracted independently, which helps to reduce redundant information between different channels. And when the high-dimensional feature map is decomposed into a series of low-dimensional feature vectors, the spatial complexity of the model can be reduced, which makes it easier for the model to learn the relationship between features, which helps to improve the learning efficiency of the model.

[0047] After the feature deconstruction is completed, the set of separated target analysis vehicle license plate color fusion group feature vectors needs to be subjected to eigenvalue-based feature decomposition to generate the corresponding set of target analysis vehicle license plate color fusion principal component eigencoding vectors. It should be understood that this step is achieved through principal component analysis (PCA). PCA is a statistical method whose theoretical basis is the eigendecomposition in linear algebra, which aims to find the main direction of data in high-dimensional space. By calculating the covariance matrix of the target analysis vehicle license plate color fusion group feature vectors and performing eigendecomposition on it, the most important direction of the data in the feature space can be found. In this way, the most representative and discriminative components in the set of original vectors can be extracted, that is, the target analysis vehicle license plate color fusion principal component eigencoding vectors. This process not only effectively reduces the feature dimension, but also compresses feature redundancy and avoids the influence of noise, which can provide a more concentrated and concise representation for subsequent data analysis.

[0048] With the help of the above principal component representation, it is necessary to further calculate the spatial distance entropy between each set of corresponding vectors in the feature vector set and the principal component intrinsic coding vector set. It should be understood that this step can quantify the spatial consistency and distribution difference between each pair of feature vectors and their principal components by combining distance measurement and entropy theory. Specifically, by calculating the spatial distance between each set of corresponding target analysis vehicle license plate color fusion feature vectors and target analysis vehicle license plate color fusion principal component intrinsic coding vectors, their relative position relationship in high-dimensional space can be evaluated. A smaller distance indicates that the two are more consistent in space. By introducing entropy, an information theory measurement method, the uniformity of the target analysis vehicle license plate color fusion feature distribution and the projection fitness of the data in the principal component space can be effectively characterized. The higher the entropy value, the more uneven the distribution or the greater the uncertainty; conversely, it means that the distribution is more concentrated or certain. By introducing spatial distance entropy as a constraint, the model parameters can be dynamically adjusted during the training process to ensure that the model focuses on the key features that best distinguish new energy vehicles from fuel vehicles. In addition, this method can also help the model better adapt to different changes in lighting, angles, etc., and improve robustness.

[0049] Next, the above-mentioned target analysis vehicle license plate color recognition space constraint vector and its transposed vector are multiplied to generate a target analysis vehicle license plate color recognition space constraint matrix. It should be understood that by multiplying the target analysis vehicle license plate color recognition space constraint vector composed of spatial distance entropy with its transposed vector, a target analysis vehicle license plate color recognition space constraint matrix can be generated, in which each element represents the association weight between feature vectors on different feature channel dimensions. This matrix is ​​essentially a quantitative expression of the global relationship in the entire target analysis vehicle license plate color recognition space. That is, the element values ​​of the target analysis vehicle license plate color recognition space constraint matrix reflect the degree of importance between feature channels and how they work together on the classification task. This helps to better understand which feature combinations are most critical for distinguishing new energy vehicles from fuel vehicles.

[0050] In order to further normalize the weight distribution in the target analysis vehicle license plate color recognition spatial constraint matrix, this matrix is ​​input into the spatial constraint activation unit based on the Sigmoid activation function to generate the activated target analysis vehicle license plate color recognition spatial constraint feature matrix. It should be understood that the element values ​​in the target analysis vehicle license plate color recognition spatial constraint matrix are normalized by a nonlinear function to strengthen the significant feature information. Specifically, the Sigmoid activation function compresses each element value in the matrix to between 0 and 1, ensuring that all features are processed on the same scale, which helps prevent certain features from dominating the model's learning process due to their large numerical range. Through this normalization process, an effective regularization effect can be achieved and the risk of overfitting can be reduced. The nonlinear compression property of the Sigmoid function can also differentially enhance the significance of features with different weights, that is, it will amplify the feature information with obvious distribution patterns or significant spatial correlations, while suppressing noise components and other unimportant features.

[0051] Finally, after obtaining the spatial constraint feature matrix of the target analysis vehicle license plate color recognition, the feature constraint operation is performed on the original target analysis vehicle license plate color fusion feature map to generate the optimized target analysis vehicle license plate color fusion feature map. It should be understood that in the spatial constraint feature matrix of the target analysis vehicle license plate color recognition, the element values ​​corresponding to the feature areas with strong significance are higher, and these high weight values ​​will further amplify the importance of the area in the feature constraint process. This means that in the optimized target analysis vehicle license plate color fusion feature map, the license plate color features for distinguishing vehicle types will be more prominent. And those areas with low relevance to the vehicle type classification task or containing noise have lower corresponding weight values, which will be reduced or even shielded in the feature constraint process, which helps to remove unnecessary complexity and interference, making the optimized feature map more concise and effective.

[0052] In step S42, the optimized target analysis vehicle license plate color fusion feature map is passed through a vehicle category classifier to obtain a vehicle classification result, and the vehicle classification result is used to indicate whether the target analysis vehicle entering is a new energy vehicle or a fuel vehicle. It should be understood that in order to accurately judge the type of vehicle based on the feature information in the optimized target analysis vehicle license plate color fusion feature map, it is necessary to input the optimized target analysis vehicle license plate color fusion feature map into the vehicle category classifier for classification processing. As a machine learning model, the vehicle category classifier can analyze and judge based on the input data and map it to different categories. The vehicle classification results obtained after classification will serve as the basis for subsequent operations. Only when the system accurately knows whether the vehicle is a new energy vehicle or a fuel vehicle can it make corresponding management decisions.

[0053] In step S43, in response to the vehicle classification result that the target analysis vehicle entering is a fuel vehicle, a voice broadcast prompt is performed and the parking lock is raised. It should be understood that in a parking lot environment, the owner may not notice the parking space logo or realize that his vehicle type does not meet the requirements of the parking space. Through voice prompts, information can be conveyed directly and effectively to prevent fuel vehicle owners from occupying new energy parking spaces without knowing it. This timely notification method can reduce parking space occupancy errors caused by information asymmetry, allowing owners to take corresponding measures immediately, such as finding a suitable ordinary parking space. The parking lock is raised to prevent fuel vehicles from entering new energy parking spaces from a physical level. Even if the owner does not hear the voice prompt or ignores the prompt information, the parking lock can prevent the vehicle from entering, thereby ensuring the exclusivity of the new energy parking space. This ensures that only new energy vehicles that meet the requirements can enter the parking space for charging and other operations, avoiding waste of resources and inconvenience to new energy vehicle owners after fuel vehicles occupy parking spaces. In general, through voice broadcast and parking lock lifting operation, fuel vehicles can be effectively prevented from occupying new energy parking spaces, so that new energy parking spaces can provide services for new energy vehicles with charging needs in a timely manner. This improves the utilization rate of new energy parking spaces, avoids idleness and waste of resources, and ensures that the charging needs of new energy vehicles can be met.

[0054] In summary, the intelligent management method of the new energy parking lot based on the embodiment of the present application is explained, which adopts the image processing technology based on computer vision, and judges whether the target vehicle is a fuel vehicle by performing high-dimensional space feature analysis on the license plate area in the image of the target vehicle. If it is a fuel vehicle, a voice broadcast prompt is performed and the parking lock is raised. In this way, the accuracy of vehicle type recognition can be significantly improved, which helps to maintain the exclusivity of new energy parking spaces in the parking lot.

[0055] Figure 5 FIG. 1 is a system block diagram of a new energy parking lot intelligent management system according to an embodiment of the present application. Figure 5 As shown, according to the embodiment of the present application, the new energy parking lot intelligent management system 100 includes: a vehicle image data acquisition module 110, which is used to acquire an image of the target analysis entering vehicle taken by a camera; a license plate area extraction module 120, which is used to perform preliminary preprocessing on the image of the target analysis entering vehicle to obtain a target analysis vehicle license plate interest region; a license plate color feature analysis module 130, which is used to analyze the target analysis vehicle license plate interest region to obtain a target analysis vehicle license plate color fusion feature map; a vehicle type response module 140, which is used to determine whether it is necessary to perform a voice broadcast prompt and execute a parking lock lifting operation according to the information in the target analysis vehicle license plate color fusion feature map.

[0056] Here, those skilled in the art can understand that the specific functions and operations of each unit and module in the new energy parking lot intelligent management system 100 have been described in the above reference. Figures 1 to 4 The description of the new energy parking lot intelligent management method has been introduced in detail, and therefore, its repeated description will be omitted.

[0057] In summary, the new energy parking lot intelligent management system 100 based on the embodiment of the present application is explained, which adopts the image processing technology based on computer vision, and judges whether the target vehicle is a fuel vehicle by performing high-dimensional space feature analysis on the license plate area in the target vehicle image. If it is a fuel vehicle, it performs voice broadcast prompts and performs parking lock lifting operation. In this way, the accuracy of vehicle type recognition can be significantly improved, which helps to maintain the exclusivity of new energy parking spaces in parking lots.

[0058] The above is only a preferred embodiment of the present application and does not constitute any form of limitation to the present application. Although the present application has been disclosed as a preferred embodiment as above, it is not intended to limit the present application. Any technician familiar with the profession can make some changes or modifications to equivalent embodiments of the technical contents disclosed above without departing from the scope of the technical solution of the present application. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present application without departing from the content of the technical solution of the present application still falls within the scope of the technical solution of the present application.

Claims

1. A new energy parking lot intelligent management method, characterized in that: include: Acquire an image of a vehicle entering the vehicle for target analysis captured by a camera; Performing preliminary preprocessing on the image of the target analysis vehicle entering the vehicle to obtain a target analysis vehicle license plate interest region; Analyze the target analysis vehicle license plate region of interest to obtain a target analysis vehicle license plate color fusion feature map; According to the target, the information in the vehicle license plate color fusion feature map is analyzed to determine whether a voice broadcast prompt and a parking lock raising operation are required.

2. The intelligent management method of new energy parking lot according to claim 1 is characterized in that: The image of the target analyzed vehicle entering the vehicle is preliminarily preprocessed to obtain the target analyzed vehicle license plate region of interest, including: Performing illumination compensation on the target analysis vehicle entering image to obtain an illumination compensated target analysis vehicle image; The illumination-compensated target analysis vehicle image is passed through a license plate target detection network to obtain the target analysis vehicle license plate region of interest.

3. The intelligent management method for new energy parking lots according to claim 2 is characterized in that: Analyzing the target analysis vehicle license plate region of interest to obtain the target analysis vehicle license plate color fusion feature map, including: Extracting the license plate color features of the target analysis vehicle license plate region of interest to obtain a target analysis vehicle license plate color feature map; Emphasizing the license plate color features of the target analysis vehicle license plate color feature map to obtain a target analysis vehicle license plate color feature enhanced map; The target analysis vehicle license plate color feature map and the target analysis vehicle license plate color feature enhancement map are fused to obtain the target analysis vehicle license plate color fusion feature map.

4. The intelligent management method for new energy parking lots according to claim 3 is characterized in that: The license plate color feature extraction is performed on the target analysis vehicle license plate interest region to obtain the target analysis vehicle license plate color feature map, including: passing the target analysis vehicle license plate interest region through a license plate color feature encoder to obtain the target analysis vehicle license plate color feature map.

5. The new energy parking lot intelligent management method according to claim 4 is characterized in that: The target analysis vehicle license plate color feature map is subjected to license plate color feature emphasis to obtain a target analysis vehicle license plate color feature enhanced map, comprising: passing the target analysis vehicle license plate color feature map through a license plate color feature enhancer to obtain the target analysis vehicle license plate color feature enhanced map.

6. The new energy parking lot intelligent management method according to claim 5 is characterized in that: The license plate color feature encoder is a convolutional neural network model as a feature extractor, and the license plate color feature enhancer is a convolutional neural network model using a channel attention mechanism.

7. The new energy parking lot intelligent management method according to claim 6 is characterized in that: According to the target, the information in the vehicle license plate color fusion feature map is analyzed to determine whether a voice broadcast prompt and a parking lock raising operation are required, including: Performing semantic region focusing optimization based on high-dimensional feature deconstruction on the target analysis vehicle license plate color fusion feature map to obtain an optimized target analysis vehicle license plate color fusion feature map; The optimized target analysis vehicle license plate color fusion feature map is passed through a vehicle category classifier to obtain a vehicle classification result, wherein the vehicle classification result is used to indicate whether the target analysis vehicle entering is a new energy vehicle or a fuel vehicle; In response to the vehicle classification result that the target analyzed incoming vehicle is a fuel vehicle, a voice broadcast prompt is performed and a parking lock raising operation is executed.

8. The new energy parking lot intelligent management method according to claim 7 is characterized in that: The target analysis vehicle license plate color fusion feature map is subjected to semantic region focusing optimization based on high-dimensional feature deconstruction to obtain an optimized target analysis vehicle license plate color fusion feature map, including: Performing feature deconstruction along the channel dimension on the target analysis vehicle license plate color fusion feature map to obtain a set of target analysis vehicle license plate color fusion feature vectors; Performing eigenvalue-based eigendecomposition on each target analysis vehicle license plate color fusion feature vector in the set of target analysis vehicle license plate color fusion feature vectors to obtain a set of target analysis vehicle license plate color fusion principal component eigencode vectors; Calculate the spatial distance entropy between each corresponding target analysis vehicle license plate color fusion feature vector and target analysis vehicle license plate color fusion principal component intrinsic coding vector in the set of the target analysis vehicle license plate color fusion feature vector and the set of the target analysis vehicle license plate color fusion principal component intrinsic coding vector to obtain a target analysis vehicle license plate color recognition spatial constraint vector composed of multiple spatial distance entropies; Calculating the product between the target analysis vehicle license plate color recognition space constraint vector and its transposed vector to obtain the target analysis vehicle license plate color recognition space constraint matrix; Inputting the target analysis vehicle license plate color recognition spatial constraint matrix into a spatial constraint activation unit based on a Sigmoid activation function to obtain a target analysis vehicle license plate color recognition spatial constraint feature matrix; Based on the target analysis vehicle license plate color recognition space constraint feature matrix, feature constraints are performed on the target analysis vehicle license plate color fusion feature map to obtain the optimized target analysis vehicle license plate color fusion feature map.

9. A new energy parking lot intelligent management system, characterized in that: include: A vehicle image data acquisition module, used to acquire an image of a vehicle entering the vehicle for target analysis captured by a camera; A license plate region extraction module is used to pre-process the image of the target analyzed vehicle to obtain a license plate region of interest of the target analyzed vehicle; A license plate color feature analysis module is used to analyze the target analysis vehicle license plate interest area to obtain a target analysis vehicle license plate color fusion feature map; The vehicle type response module is used to analyze the information in the vehicle license plate color fusion feature map according to the target, and determine whether it is necessary to perform a voice broadcast prompt and execute the parking lock raising operation.

10. The new energy parking lot intelligent management system according to claim 9 is characterized in that: The license plate area extraction module comprises: An image illumination compensation unit, used for performing illumination compensation on the target analysis vehicle image to obtain an illumination compensated target analysis vehicle image; The vehicle license plate region of interest extraction unit is used to pass the illumination compensated target analysis vehicle image through a license plate target detection network to obtain the target analysis vehicle license plate region of interest.

Images