An intelligent parking lot management system for a smart community

Abstract

The present invention relates to the field of intelligent management technologies, and provides an intelligent parking lot management system for a smart community, which includes a vehicle type recognition module, a real-time monitoring module, a dynamic adjustment module, and an intelligent guidance module. The vehicle type recognition module uses lidar sensor technology to classify vehicle types according to the vehicle length. The parking space detection module monitors the occupancy of parking spaces in the parking lot in real time, collects image samples and uploads them to the dynamic adjustment module. The dynamic adjustment module uses a convolutional neural network to analyze the occupancy of parking spaces, predict the full-load probability of the parking lot, and adjust the parking space layout accordingly. The intelligent guidance module provides parking space guidance for drivers based on the vehicle type recognition results of the vehicle type recognition module and the adjustment results of the dynamic adjustment module.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent management, and more specifically, to an intelligent parking lot management system for a smart community. Background Art

[0002] With the sharp increase in the number of car purchases, people's requirements for parking lots are also constantly increasing. When planning parking spaces, traditional parking management systems usually only consider the largest-sized vehicles, resulting in small and compact vehicles having difficulty making full use of the overly large parking spaces, causing waste of the parking lot's space resources. Also, due to the inflexibility of the parking space areas, when the parking lot needs to accommodate different types and sizes of vehicles, there is often a shortage of parking spaces. Moreover, because vehicles stay for different lengths of time, if vehicles are parked according to the capacity for the largest vehicles, it will cause inconvenience for vehicles to enter and exit. Community parking lots need to facilitate vehicle entry and exit while saving space, and at this time, dynamic intervention in the vehicles in the parking lot is required. Summary of the Invention

[0003] To overcome the above-mentioned defects of the prior art, the present invention provides an intelligent parking lot management system for a smart community. The dynamic adjustment module uses a convolutional neural network to analyze the occupancy of parking spaces, predict the full-load probability of the parking lot, and adjust the parking space layout accordingly. The intelligent guidance module provides parking space guidance for the driver based on the vehicle type recognition result of the vehicle type recognition module and the adjustment result of the dynamic adjustment module to help the driver reach the assigned parking space.

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] An intelligent parking lot management system for a smart community, comprising a vehicle type recognition module, a parking space detection module, a dynamic adjustment module, and an intelligent guidance module. The vehicle type recognition module is used to measure the length of the vehicle driving into the parking lot in real time through lidar sensor technology and classify the vehicles according to their lengths. The parking space detection module is used to collect real-time parking lot image samples, detect the occupancy of parking spaces in the parking lot, and upload the detection results to the dynamic adjustment module. The dynamic adjustment module is used to analyze the occupancy of parking spaces using a convolutional neural network, predict the full-load probability of the parking lot, adjust the parking space layout of the parking lot when the full-load probability reaches a threshold, and upload the adjustment results to the intelligent guidance module. The intelligent guidance module is used to provide parking space guidance for the driver based on the vehicle type and the adjustment results of the dynamic management module; the intelligent guidance module is set on the display screens at the entrance of the parking lot and at the entrance of each area.

[0006] As a further solution of the present invention, the image acquisition unit, the image processing unit, the parking space status detection unit, and the result output unit are all connected to the parking space detection module. The image acquisition unit is connected to the image processing unit, the image processing unit is connected to the parking space status detection unit, the parking space status detection unit is connected to the result output unit. The full-load prediction unit and the parking space adjustment unit are both connected to the dynamic adjustment module, the full-load prediction unit is connected to the parking space adjustment unit, the parking space detection module is connected to the dynamic adjustment module, and the vehicle type identification module and the dynamic adjustment module are both connected to the intelligent guidance module.

[0007] As a further solution of the present invention, the vehicle type identification module classifies vehicles with a length of less than 3700 mm as mini cars, vehicles with a length of 3700 - 4300 mm as small cars, vehicles with a length of 4300 - 4600 mm as compact cars, vehicles with a length of 4600 - 4900 mm as medium-sized cars, vehicles with a length of 4900 - 5100 mm as large medium-sized cars, and vehicles with a length of more than 5100 mm as large cars.

[0008] As a further solution of the present invention, the parking lot is divided into Area A, Area B, and Area C. A site with an area accounting for 20% of the area of Area A is set around the entrance of Area A as a temporary parking area, and the remaining 80% of the site in Area A is a regular parking area. The parking lines in the temporary parking area are planned as vertical parking spaces using LED light strips and double lines. The parking lines in the regular parking area are marked as 45° inclined parking spaces. The regular parking area in Area A is a dedicated parking area for mini cars and small cars, and the temporary parking area in Area A is a secondary parking area for mini cars, small cars, compact cars, and medium-sized cars. Area B is a dedicated parking area for compact cars and medium-sized cars, and Area C is a dedicated parking area for large medium-sized cars and large cars. Among them, the specific steps for planning the parking lines in the temporary parking area using LED light strips and double lines as vertical parking spaces are as follows:

[0009] Step A1, determination of parking space size: Set the vertical parking space size for mini cars and small cars in the temporary parking area as 2400×5500 mm, and the vertical parking space size for compact cars and medium-sized cars as 2600×6000 mm;

[0010] Step A2, installation of LED light strips: Install LED light strips showing white light at the edges of the parking spaces for mini cars and small cars set in the temporary parking area of Area A, and install LED light strips showing yellow light at the edges of the parking spaces for compact cars and medium-sized cars set in the temporary parking area of Area A. There is no overlapping part when installing the two types of light strips;

[0011] Step A3, System Integration: Integrate the brightness adjustment and dynamic adjustment module of the LED light strip, and adjust the startup and shutdown of the white LED light strip and the yellow LED light strip according to the adjustment results of the dynamic adjustment module.

[0012] As a further solution of the present invention, the parking space detection module includes an image acquisition unit, an image processing unit, a parking space status detection unit, and a result output unit. The image acquisition unit captures real-time image samples within the entire parking lot by placing 1080P cameras at the entrance and exit of the parking lot, above the parking spaces, and beside the lanes; the image processing unit preprocesses the acquired image samples, including image correction, denoising, grayscale conversion, and normalization operations; the parking space status detection unit uses object recognition technology to identify the parking spaces and vehicles in the image samples, and detects the occupancy of the parking spaces according to the recognition results. If a vehicle is recognized in a parking space, the parking space is marked as occupied; if no vehicle is recognized in the parking space, the parking space is marked as free; the result output unit uploads the image marked with the occupancy of the parking spaces output by the parking space detection unit to the dynamic adjustment module.

[0013] As a further solution of the present invention, the dynamic adjustment module includes a full-load prediction unit and a parking space adjustment unit. The full-load prediction unit uses a convolutional neural network to analyze the output image marked with the occupancy of the parking spaces uploaded by the parking space detection module to predict the full-load probability of Parking Lot B; the parking space adjustment unit adjusts the parking space layout in Area A according to the prediction results of the full-load prediction unit. When the full-load probability in Area B is lower than 70%, set the regular parking area in Area A as the priority parking area for compact cars and small cars, and set the temporary parking area in Area A as the secondary priority parking area for compact cars and small cars. The white LED light strip on the parking space in the temporary parking area is turned on, and the yellow LED light strip is turned off. At this time, the non-overlapping area inside the parking spaces demarcated by the white LED light strip and the yellow LED light strip will be used as a passage. When the full-load probability in Area B is greater than or equal to 70%, the parking space adjustment unit will automatically cancel the secondary priority parking qualification of compact cars and small cars in the temporary parking area of Area A. At this time, this temporary parking area is specifically used as the parking area for mid-size cars and large cars, while compact cars and small cars are only allowed to park in the regular parking area of Area A. The white LED light strip on the parking space in the temporary parking area is turned off, and the yellow LED light strip is turned on. The non-overlapping area inside the parking spaces demarcated by the white LED light strip and the yellow LED light strip in the original temporary parking area of Area A that was used as a passage is converted into a parking space for vehicles to use. Among them, the steps of predicting the full-load probability of Parking Lot B using a convolutional neural network are as follows:

[0014] Step 1, Determine the input layer: Receive the image marked with the occupancy of the parking spaces uploaded by the parking space detection module as the input layer;

[0015] Step 2, determine the convolutional layer: The convolutional layer uses a 7×7 convolutional kernel to slide on the input image to identify the overall contour features of the vehicle and generate a feature map, uses a 3×3 convolutional kernel to slide on the input image to identify the parking space boundary line features and generate a feature map, uses a 5×5 L-shaped convolutional kernel to slide on the image to identify the arrangement and density features of the vehicles and generate a feature map, captures the spatial hierarchical structure of the above feature maps, performs convolution calculations on the pixel values at the pixel positions where the convolutional kernel slides on the input image, obtains the feature values at the pixel positions on the feature map, and then applies an activation function to each pixel on the above feature map to introduce non-linear features. Among them, the formula for obtaining the feature values at the pixel positions on the feature map is:

[0016]

[0017] In the formula: F(x,y) is the feature value at the (x,y) position on the feature map, m, n are the sizes of the convolutional kernel, x is the horizontal axis coordinate of the pixel position on the input image, y is the vertical axis coordinate of the pixel position on the input image, I(x,y) is the pixel value of the input image at the position (x,y), and K(x,y) is the weight value of the convolutional kernel at (x,y);

[0018] The formula for the activation function of the convolutional layer is:

[0019]

[0020] In the formula: F(x,y) is the feature value at the (x,y) position on the feature map, and A(x,y) is the activation feature value at the (x,y) position output by the convolutional layer;

[0021] Step 3, determine the pooling layer: During the pooling process, the feature map after applying the activation function is divided into non-overlapping pooling windows of size 2X2 with a stride of 2, and pooling operations are performed on each pooling window to obtain the pooling values at the pixel positions in the pooling window. Among them, the formula for obtaining the pooling values at the pixel positions in the pooling window is:

[0022]

[0023] In the formula: P(x,y) is the pooling value at the (x,y) position in the pooling window;

[0024] Step 4, determine the fully connected layer: Flatten the output result of the pooling layer, convert the output result of the pooling layer from a two-dimensional vector to a one-dimensional vector as the input of the fully connected layer, then calculate the weighted sum of this one-dimensional vector, pass the obtained weighted sum to the activation function for non-linear mapping, integrate the information in all feature maps, and obtain the output result of the fully connected layer. This output result is a new one-dimensional vector. Among them, the formula for calculating the weighted sum is:

[0025]

[0026] Where: C is the weighted sum of the one-dimensional vector input to the fully connected layer, n is the length of the one-dimensional vector, k is the index of each element in the one-dimensional vector connected to the fully connected layer, P[k] is the k-th element of the one-dimensional vector, V(k) is the weight of P(k) connected to the fully connected layer, and b is the bias term;

[0027] The formula for the activation function of the fully connected layer is:

[0028]

[0029] Where: G(C) is the output result of the fully connected layer;

[0030] Step Five, determine the output layer: Determine that the number of neurons in the output layer is set to 1. Apply the activation function to the output result of the fully connected layer to convert the output result of the fully connected layer into a probability value, indicating the possibility of full load in Area B. Among them, the formula for the activation function of the output layer is:

[0031]

[0032] Where: P out is the probability value of full load in Area B, and γ is the scaling parameter.

[0033] As a further solution of the present invention, the intelligent guidance module receives the vehicle type recognition result of the vehicle type recognition module, combines the available parking spaces recognized by the parking space detection module and the adjustment result of the dynamic adjustment module, and guides sedans of different levels to the specified positions. When the full load probability of Area B is less than 70%, mini-cars and small cars are preferentially guided to the regular parking area of Area A, while the temporary parking area is used as the secondary preferential parking area for mini-cars and small cars. Compact cars and medium-sized cars are guided to the parking area of Area B, and large and medium-sized cars and large cars are guided to the parking area of Area C. When the full load probability of Area B is greater than or equal to 70%, mini-cars and small cars are guided to the regular parking area of Area A, compact cars and medium-sized cars are preferentially guided to the parking area of Area B, while the temporary parking area of Area A is used as the secondary preferential parking area for compact cars and medium-sized cars, and large and medium-sized cars and large cars are guided to the parking area of Area C. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 is a schematic structural diagram of an intelligent parking lot management system for a smart community according to the present invention;

[0035] Figure 2 is a schematic flow diagram of an intelligent parking lot management system for a smart community according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0036] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0037] Embodiment 1

[0038] As Figure 1 shown, an intelligent parking lot management system for a smart community includes a vehicle type recognition module, a parking space detection module, a dynamic adjustment module, and an intelligent guidance module. The vehicle type recognition module is used to measure the length of the vehicle driving into the parking lot in real time through lidar sensor technology and classify the vehicles according to their lengths. In the embodiments of the present invention, the vehicle type recognition module classifies vehicles with a length of less than 3700 mm as mini cars, vehicles with a length of 3700 - 4300 mm as small cars, vehicles with a length of 4300 - 4600 mm as compact cars, vehicles with a length of 4600 - 4900 mm as medium-sized cars, vehicles with a length of 4900 - 5100 mm as medium and large-sized cars, and vehicles with a length of more than 5100 mm as large cars. The parking lot in the embodiments of the present invention is divided into Area A, Area B, and Area C, and a site with an area of 20% of the area of Area A is set around the entrance of Area A as a temporary parking area, and the remaining 80% of the site in Area A is a regular parking area. The temporary parking area uses LED light strips and double lines to design the parking lines in a vertical parking space. The parking lines in the regular parking area are marked with 45° inclined parking spaces. The regular parking area in Area A is a dedicated parking area for mini cars and small cars, and the temporary parking area in Area A is a secondary parking area for mini cars, small cars, compact cars, and medium-sized cars. Area B is a dedicated parking area for compact cars and medium-sized cars, and Area C is a dedicated parking area for medium and large-sized cars and large cars. Among them, the specific steps for using LED light strips and double lines to design the parking lines in the vertical parking space of the temporary parking area are as follows:

[0039] Step A1, determination of parking space size: Set the vertical parking space size for mini cars and small cars in the temporary parking area to 2400×5500 mm, and the vertical parking space size for compact cars and medium-sized cars to 2600×6000 mm.

[0040] Step A2, installation of LED light strips: Install LED light strips showing white light at the edges of the parking spaces for mini cars and small cars set in the temporary parking area of Area A, and install LED light strips showing yellow light at the edges of the parking spaces for compact cars and medium-sized cars set in the temporary parking area of Area A. There is no overlapping part when installing the two types of light strips.

[0041] Step A3, System Integration: Integrate the brightness adjustment of the LED light strips with the dynamic adjustment module, and adjust the startup and shutdown of the white LED light strips and the yellow LED light strips according to the adjustment results of the dynamic adjustment module;

[0042] The parking space detection module is used to collect real-time parking lot image samples, detect the occupancy of parking spaces in the parking lot, and upload the detection results to the dynamic adjustment module;

[0043] The parking space detection module in the embodiments of the present invention includes an image acquisition unit, an image processing unit, a parking space status detection unit, and a result output unit. The image acquisition unit captures real-time image samples within the entire parking lot by placing 1080P cameras at the entrance and exit of the parking lot, above the parking spaces, and beside the lanes; the image processing unit preprocesses the collected image samples, including image correction, denoising, grayscale conversion, and normalization operations; the parking space status detection unit uses object recognition technology to identify the parking spaces and vehicles in the image samples, and detects the occupancy of the parking spaces according to the recognition results. If a vehicle is recognized in a parking space, the parking space is marked as occupied. If no vehicle is recognized in the parking space, the parking space is marked as free; the result output unit uploads the image marked with the occupancy of the parking spaces output by the parking space detection unit to the dynamic adjustment module. Among them, the specific steps of using object recognition technology to identify the parking spaces and vehicles in the image samples are as follows:

[0044] Step S1, Background Modeling: Select the internal image of the parking lot taken during a period without vehicles as the background model;

[0045] Step S2, Background Subtraction: Subtract the background model from the real-time image samples in the parking lot, leaving only the objects in the image;

[0046] Step S3, Contour Segmentation: Use the Canny edge detector to find the edges of the objects in the image, thereby determining the object contours, and separating the objects from the background according to the object contours;

[0047] Step S4, Object Size and Shape Analysis: Analyze the size and shape of the detected objects to determine whether they conform to the characteristics of vehicles;

[0048] Step S5, Position Judgment: Judge whether the object is within the parking space area according to the position of the object in the image

[0049] Step S6, Status Judgment: If the detected object conforms to the characteristics of a vehicle and is located within the parking space area, then judge that the parking space is "occupied", otherwise, judge it as "free";

[0050] The dynamic adjustment module is used to analyze the occupancy of parking spaces using a convolutional neural network, predict the full-load probability of the parking lot, adjust the parking space layout of the parking lot when the full-load probability reaches the threshold, and upload the adjustment results to the intelligent guidance module;

[0051] The dynamic adjustment module in the embodiments of the present invention includes a full-load prediction unit and a parking space adjustment unit. The full-load prediction unit analyzes the output image marked with the occupancy situation of parking spaces uploaded by the parking space detection module using a convolutional neural network to predict the full-load probability of the parking lot in Area B. The parking space adjustment unit adjusts the parking space layout in Area A according to the prediction result of the full-load prediction unit. When the full-load probability in Area B is lower than 70%, the regular parking area in Area A is set as the priority parking area for compact cars and small cars, and the temporary parking area in Area A is set as the secondary priority parking area for compact cars and small cars. The white LED strip on the parking space in the temporary parking area is activated, and the yellow LED strip is turned off. At this time, the non-overlapping area inside the parking space demarcated by the white LED strip and the yellow LED strip will be used as a passage. When the full-load probability in Area B is greater than or equal to 70%, the parking space adjustment unit will automatically cancel the secondary priority parking qualification of compact cars and small cars in the temporary parking area of Area A. At this time, this temporary parking area is specifically used as the parking area for mid-size cars and large cars, while compact cars and small cars are only allowed to park in the regular parking area of Area A. The white LED strip on the parking space in the temporary parking area is turned off, and the yellow LED strip is activated. The non-overlapping area inside the parking space demarcated by the white LED strip and the yellow LED strip in the original temporary parking area of Area A used as a passage is converted into a parking space for vehicles to use. Among them, the steps of predicting the full-load probability of the parking lot in Area B using a convolutional neural network are as follows:

[0052] Step 1, determine the input layer: Receive the image marked with the occupancy situation of parking spaces uploaded by the parking space detection module as the input layer;

[0053] Step 2, determine the convolutional layer: The convolutional layer uses a 7×7 convolutional kernel to slide on the input image to identify the overall contour features of vehicles and generate a feature map, uses a 3×3 convolutional kernel to slide on the input image to identify the boundary line features of parking spaces and generate a feature map, uses a 5×5 L-shaped convolutional kernel to slide on the image to identify the arrangement and density features of vehicles and generate a feature map, captures the spatial hierarchical structure of the above feature maps, performs convolutional calculations on the pixel values at the pixel positions where the convolutional kernel slides on the input image to obtain the feature values at the pixel positions on the feature map. Subsequently, an activation function is applied to each pixel on the above feature map to introduce non-linear features. Among them, the formula for obtaining the feature value at the pixel position on the feature map is:

[0054]

[0055] In the formula: F(x,y) is the feature value at the position (x,y) on the feature map, m, n are the sizes of the convolutional kernel, x is the horizontal axis coordinate of the pixel position of the input image, y is the vertical axis coordinate of the pixel position of the input image, I(x,y) is the pixel value of the input image at the position (x,y), and K(x,y) is the weight value of the convolutional kernel at (x,y);

[0056] The formula for the activation function of the convolutional layer is as follows:

[0057]

[0058] Where: F(x, y) is the feature value at the position (x, y) on the feature map, and A(x, y) is the activation feature value at the position (x, y) output by the convolutional layer;

[0059] Step 3, determine the pooling layer: During the pooling process, the feature map after applying the activation function is divided into non-overlapping pooling windows of size 2X2 with a stride of 2, and pooling operations are performed on each pooling window to obtain the pooling value at the pixel position in the pooling window. The formula for obtaining the pooling value at the pixel position in the pooling window is as follows:

[0060]

[0061] Where: P(x, y) is the pooling value at the position (x, y) in the pooling window;

[0062] Step 4, determine the fully connected layer: Flatten the output result of the pooling layer, convert the output result of the pooling layer from a two-dimensional vector to a one-dimensional vector as the input of the fully connected layer. Subsequently, calculate the weighted sum of this one-dimensional vector, pass the obtained weighted sum to the activation function for non-linear mapping, and integrate the information in all feature maps to obtain the output result of the fully connected layer. This output result is a new one-dimensional vector. The formula for calculating the weighted sum is as follows:

[0063]

[0064] Where: C is the weighted sum of the one-dimensional vector input to the fully connected layer, n is the length of the one-dimensional vector, k is the index of the connection between each element in the one-dimensional vector and the fully connected layer, P[k] is the k-th element of the one-dimensional vector, V(k) is the weight of the connection between P(k) and the fully connected layer, and b is the bias term;

[0065] The formula for the activation function of the fully connected layer is as follows:

[0066]

[0067] Where: G(C) is the output result of the fully connected layer;

[0068] Step 5, determine the output layer: Determine that the number of neurons in the output layer is set to 1. Apply the activation function to the output result of the fully connected layer to convert the output result of the fully connected layer into a probability value indicating the possibility of full load in area B. The formula for the activation function of the output layer is as follows:

[0069]

[0070] Where: P out is the probability value of full load in Area B, and γ is the scaling parameter;

[0071] The intelligent guidance module is used to provide parking space guidance for the driver according to the vehicle type and the adjustment result of the dynamic management module;

[0072] In the embodiment of the present invention, the intelligent guidance module receives the vehicle type recognition result of the vehicle type recognition module, combines the available parking spaces recognized by the parking space detection module and the adjustment result of the dynamic adjustment module, and guides sedans of different levels to the specified positions. When the full load probability in Area B is lower than 70%, mini-cars and small cars are preferentially guided to the regular parking areas in Area A, while the temporary parking areas are the secondary preferential parking areas for mini-cars and small cars. Compact cars and medium-sized cars are guided to the parking areas in Area B, and large and medium-sized cars and large cars are guided to the parking areas in Area C. When the full load probability in Area B is greater than or equal to 70%, mini-cars and small cars are guided to the regular parking areas in Area A, compact cars and medium-sized cars are preferentially guided to the parking areas in Area B, while the temporary parking areas in Area A are the secondary preferential parking areas for compact cars and medium-sized cars, and large and medium-sized cars and large cars are guided to the parking areas in Area C; The intelligent guidance module is set on the display screens at the entrance of the parking lot and at the entrances of each area.

[0073] The following is a simple Python code example that uses the DT algorithm to provide real-time intelligent guidance for vehicles. Please note that this example is only a starting point and may need to be adjusted according to the actual situation and device interfaces in actual applications;

[0074] First, dock the API interface:

[0075]

[0076]

[0077]

[0078] This code defines a simple intelligent guidance module that uses the DT algorithm to provide real-time intelligent guidance for vehicles. Please adjust the code according to the actual device interfaces and requirements.

[0079] In the embodiments of the present invention, the image acquisition unit, the image processing unit, the parking space status detection unit, and the result output unit are all connected to the parking space detection module. The image acquisition unit is connected to the image processing unit, the image processing unit is connected to the parking space status detection unit, and the parking space status detection unit is connected to the result output unit. The full-load prediction unit and the parking space adjustment unit are both connected to the dynamic adjustment module. The full-load prediction unit is connected to the parking space adjustment unit. The parking space detection module is connected to the dynamic adjustment module. The vehicle type identification module and the dynamic adjustment module are both connected to the intelligent guidance module.

[0080] Embodiment 2

[0081] Figure 2 It is a schematic flowchart of an intelligent parking lot management system for a smart community. When a vehicle enters the parking lot, first, the length of the vehicle is measured in real time through lidar sensor technology, and the vehicle is classified according to the length. At the same time, the system collects real-time image samples of the parking lot and detects the occupancy of parking spaces. Then, convolutional neural network is used to analyze the parking space occupancy data to predict the full-load probability of the parking area in Zone B. If the predicted full-load probability is greater than or equal to the set full-load probability threshold, the system preferentially guides mini-cars and small cars to the regular parking area in Zone A, and the temporary parking area is the secondary priority parking area. The white LED light strips on the parking spaces are activated, and the yellow LED light strips are turned off. At the same time, compact cars and medium-sized cars are guided to the parking area in Zone B, while large and medium-sized cars and large cars are guided to the parking area in Zone C. If the full-load probability is lower than the threshold, mini-cars and small cars are guided to the regular parking area in Zone A, and compact cars and medium-sized cars are preferentially guided to Zone B. The temporary parking area in Zone A is the secondary priority parking area for compact cars and medium-sized cars. The white LED light strips on the parking spaces are turned off, and the yellow LED light strips are activated. Large and medium-sized cars and large cars are still guided to Zone C, and the process ends.

[0082] As mentioned above, it is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed by this application can easily think of changes or substitutions, which should all be covered within the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

[0083] Finally: As mentioned above, it is only the preferred embodiment of the present invention and is not used to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should all be included within the protection scope of the present invention.

Claims

1. A smart community intelligent parking lot management system, characterized in that: It includes a vehicle type recognition module, a parking space detection module, a dynamic adjustment module and an intelligent guidance module; the parking space detection module includes an image acquisition unit, an image processing unit, a parking space state detection unit and a result output unit, the image acquisition unit is connected to the image processing unit, the image processing unit is connected to the parking space state detection unit, the parking space state detection unit is connected to the result output unit, the dynamic adjustment module includes a full load prediction unit and a parking space adjustment unit, the full load prediction unit is connected to the parking space adjustment unit, the parking space detection module is connected to the dynamic adjustment module, and the vehicle type recognition module and the dynamic adjustment module are jointly connected to the intelligent guidance module; The vehicle type recognition module is used to measure the length of vehicles entering the parking lot in real time through a laser radar sensor and classify the vehicles according to their length; The parking space detection module is used to collect parking lot image samples in real time, detect the parking space occupancy in the parking lot, and upload the detection results to the dynamic adjustment module; the dynamic adjustment module is used to use convolutional neural networks to analyze the parking space occupancy and predict the probability of full load of the parking lot. When the full load probability reaches the threshold, the parking space layout of the parking lot is adjusted, and the adjustment results are uploaded to the intelligent guidance module; The intelligent guidance module is used to provide parking guidance to the driver based on the vehicle type and the adjustment results of the dynamic management module; The management system specifically includes: dividing the parking lot into Area A, Area B and Area C, and setting up a site with an area of ​​20% of the area of ​​Area A around the entrance of Area A as a temporary parking area, and the remaining 80% of the site in Area A is a regular parking area. The temporary parking area uses LED light bars and double-line design to plan parking lines with vertical parking spaces, and the parking space lines in the regular parking area are marked with 45° inclined parking spaces. The regular parking area in Area A is a special parking area for micro cars and small cars, and the temporary parking area in Area A is a secondary parking area for micro cars, small cars, compact cars and medium-sized cars. Area B is a special parking area for compact cars and medium-sized cars, and Area C is a special parking area for medium and large cars and large cars. Among them, the specific steps of using LED light bars and double-line design to plan parking lines in the temporary parking area with vertical parking spaces are as follows: Step A1, determining the parking space size: setting the vertical parking space size of mini cars and small cars in the temporary parking area to 2400×5500mm, and the vertical parking space size of compact cars and mid-sized cars to 2600×6000mm; Step A2, installing LED light bars: installing LED light bars displaying white light at the edge of parking spaces for mini cars and small cars set in the temporary parking area of ​​Zone A, and installing LED light bars displaying yellow light at the edge of parking spaces for compact cars and mid-sized cars set in the temporary parking area of ​​Zone A, and the two light bars are installed without overlapping parts; Step A3, system integration: Integrate the brightness adjustment of the LED light bar with the dynamic adjustment module, and adjust the start and stop of the white LED light bar and the yellow LED light bar according to the adjustment result of the dynamic adjustment module.

2. According to claim 1, a smart community intelligent parking lot management system is characterized in that: The vehicle type identification module classifies vehicles with a length of less than 3700mm as micro cars, vehicles with a length of 3700-4300mm as small cars, vehicles with a length of 4300-4600mm as compact cars, vehicles with a length of 4600-4900mm as mid-size cars, vehicles with a length of 4900-5100mm as mid-to-large cars, and vehicles with a length of more than 5100mm as large cars.

3. According to claim 1, a smart community intelligent parking lot management system is characterized in that: The full load prediction unit uses a convolutional neural network to analyze the output image marked with parking space occupancy uploaded by the parking space detection module, and predicts the full load probability of the parking lot in area B; the parking space adjustment unit adjusts the parking space layout in area A according to the prediction result of the full load prediction unit. When the full load probability of area B is lower than 70%, the regular parking area in area A is set as the priority parking area for micro cars and small cars, and the temporary parking area in area A is set as the second priority parking area for micro cars and small cars. The white LED light bar on the parking space in the temporary parking area is turned on and the yellow LED light bar is turned off. At this time, the non-overlapping area inside the parking space demarcated by the white LED light bar and the yellow LED light bar will be used as an aisle. When B When the probability of full load in the area is greater than or equal to 70%, the parking space adjustment unit will automatically cancel the second priority parking qualification of micro cars and small cars in the temporary parking area of ​​area A. At this time, the temporary parking area is specially used as a parking area for compact cars and medium-sized cars, while micro cars and small cars will be limited to parking in the regular parking area of ​​area A. The white LED light bar on the parking space in the temporary parking area is turned off, and the yellow LED light bar is turned on. The non-overlapping area inside the parking space demarcated by the white LED light bar and the yellow LED light bar in the temporary parking area of ​​area A, which was originally used as an aisle, is converted into a parking space for vehicles. Among them, the steps of using convolutional neural network to predict the probability of full load of the parking lot in area B are as follows: Step 1, determining the input layer: receiving the image uploaded by the parking space detection module and marked with the parking space occupancy status as the input layer; Step 2, determine the convolution layer: the convolution layer uses a 7×7 convolution kernel to slide on the input image to identify the overall contour features of the vehicle and generate a feature map, uses a 3×3 convolution kernel to slide on the input image to identify the parking space boundary line features and generate a feature map, and uses a 5×5 L-shaped convolution kernel to slide on the image to identify the arrangement and density features of the vehicle and generate a feature map. The spatial hierarchical structure of the above feature map is captured, and the pixel value of each convolution kernel sliding position on the input image is convoluted to obtain the feature value on the feature map. Subsequently, an activation function is applied to each pixel of the above feature map to introduce nonlinear features. The formula for convolution calculation is: ; Where: For the feature map The eigenvalues ​​of the positions, , is the size of the convolution kernel, is the horizontal axis coordinate of the input image, is the vertical coordinate of the input image, For the input image at position The pixel value of The convolution kernel is The weight value at ; The formula for the convolutional layer activation function is: ; Where: For the feature map The eigenvalues ​​of the positions, The output of the convolutional layer The activation feature value of the position; Step 3: Determine the pooling layer: During the pooling process, the feature map after the activation function is applied is divided into non-overlapping pooling windows of size 2X2 and stride 2, and a pooling operation is performed on each pooling window. The formula for the pooling operation is: ； Where: In the pooling window Pooled value of position; Step 4: Determine the fully connected layer: Flatten the output of the pooling layer, transform the output of the pooling layer from a two-dimensional vector to a one-dimensional vector as the input of the fully connected layer, then calculate the weighted sum of the one-dimensional vector, pass the obtained weighted sum to the activation function, perform nonlinear mapping, integrate the information in all feature maps, and obtain the output of the fully connected layer. The output result is a new one-dimensional vector, where the formula for calculating the weighted sum is: ； Where: is the weighted sum of the one-dimensional vector input to the fully connected layer, is the length of a one-dimensional vector, is the index of each element in the one-dimensional vector connected to the fully connected layer, is the kth element of a one-dimensional vector, for The weights of the connections to the fully connected layers, is the bias term; The formula of the fully connected layer activation function is: ; Where: is the output result of the fully connected layer; Step 5: Determine the output layer: Determine the number of neurons in the output layer and set it to 1. Apply the activation function to the output of the fully connected layer to convert the output of the fully connected layer into a probability value, indicating the possibility that area B is fully loaded. The formula of the output layer activation function is: ； Where: is the probability value of area B being fully loaded, is the scaling parameter.

4. According to claim 1, a smart community intelligent parking lot management system is characterized in that: The image acquisition unit captures real-time image samples of the entire parking lot by placing 1080p cameras at the entrance and exit of the parking lot, above the parking spaces, and beside the lanes; the image processing unit pre-processes the collected image samples, including image correction, denoising, grayscale, and normalization operations; The parking space status detection unit uses object recognition technology to identify parking spaces and vehicles in image samples, and detects the parking space occupancy status based on the recognition results. If a vehicle is recognized in the parking space, the parking space is marked as occupied. If no vehicle is recognized in the parking space, the parking space is marked as free. The result output unit uploads the image output by the parking space detection unit and marked with the parking space occupancy status to the dynamic adjustment module.

5. The smart community intelligent parking lot management system according to claim 1 is characterized in that: The intelligent guidance module receives the vehicle type recognition result of the vehicle type recognition module, combines the vacant parking spaces identified by the parking space detection module and the adjustment result of the dynamic adjustment module, and guides cars of different levels to the specified positions. When the probability of full load in area B is less than 70%, micro cars and small cars are guided to the regular parking area in area A, and the temporary parking area is used as the second priority parking area for micro cars and small cars. Compact cars and medium-sized cars are guided to the parking area in area B, and medium-sized and large cars are guided to the parking area in area C. When the probability of full load in area B is greater than or equal to 70%, micro cars and small cars are guided to the regular parking area in area A, compact cars and medium-sized cars are guided to the parking area in area B, and the temporary parking area in area A is used as the second priority parking area for compact cars and medium-sized cars, and medium-sized and large cars are guided to the parking area in area C.

Images