Parking space guiding management system and method based on unmanned aerial vehicle management

Abstract

The invention discloses a parking space guidance management system and method based on unmanned aerial vehicle management, and relates to the technical field of parking space guidance management, and the system comprises an image acquisition unit, an edge analysis unit, a parking line recognition unit, a parking space coordinate determination unit, a parking space state extraction unit and a path analysis unit. The parking line identification unit is used for analyzing the edge points to obtain all parking lines in the image information, so that the parking space coordinate determination unit can calculate the specific position of the corresponding parking space according to the position of the parking line and the width of the parking space, and the mode of determining the coordinates of the parking space saves a large amount of manpower; moreover, the accurate position of each parking space can be determined in a short time, and the parking path of each to-be-parked vehicle and the conflict probability corresponding to each parking path are obtained through a path analysis unit according to the idle parking spaces recognized by the parking space state extraction unit, so that on one hand, the unmanned aerial vehicle is ensured to complete a guiding task; and on the other hand, the possible blockage probability is predicted.

Description

Technical Field

[0001] The present invention relates to the technical field of parking space guidance management, and in particular to a parking space guidance management system and method based on unmanned aerial vehicle management. Background Art

[0002] The parking guidance system using drones can effectively guide and manage parked vehicles entering the parking lot. This system can enable parkers to park conveniently and quickly, making the parking space management of the parking lot more standardized and orderly. The invention patent with application number CN202010410417.0 discloses "a parking guidance system and working method based on RFID and drones, which combines RFID technology with software radio and applies it to the field of parking. The radio frequency signal of the RFID radio license plate is obtained by carrying the USRPB200 device on the drone, and then the vehicle position is accurately located after being processed by the RFID indoor positioning algorithm. The camera on the drone collects the environmental information and parking space information around the vehicle and sends it to the management platform, which is analyzed and processed by the image processing module of the management platform to obtain the location of the unselected vacant parking space. The vehicle position, the location of the vacant parking space and the environmental information around the vehicle are used as input parameters for route planning, and the shortest parking route is planned for the vehicle through the route planning algorithm. Quickly help car owners find the nearest vacant parking space to the vehicle, and the drone guides parking according to the planned route, saving the parking time of the car owner."

[0003] The above-mentioned existing technology solves the problem of difficulty in finding parking spaces for vehicles. However, during the use of the system, since the edge points are not repeatedly screened during the image analysis process, large errors may occur when identifying the parking line, making it impossible to subsequently obtain the more accurate specific location of the parking space. In order to obtain more accurate location information, too much equipment needs to be used, which increases the cost to a certain extent. In addition, the system cannot predict the probability of congestion when formulating a parking route. Summary of the invention

[0004] The purpose of the present invention is to provide a parking space guidance management system and method based on drone management to solve the problems raised in the above background technology.

[0005] To achieve the above-mentioned object, the present invention provides the following technical solutions: a parking space guidance management system based on drone management, comprising a parking space status extraction unit and a path analysis unit;

[0006] An image acquisition unit, wherein the image acquisition unit extracts the total number of parking spaces in the parking lot and the number of currently parked vehicles. If the ratio between the number of currently parked vehicles and the total number of parking spaces is less than 0.15, the OV13850 camera carried by the drone is used to shoot the entire parking lot at a preset flight altitude, and the acquired image information is transmitted to a storage device;

[0007] An edge analysis unit, after acquiring the collected image information, selects a 3×3 filter to sort each pixel in the image information and its surrounding adjacent pixels according to grayscale values, uses the median value of the sequence as the grayscale value of the pixel, calculates the gradient amplitude and direction of each pixel using a Soble convolution template, determines the initial edge point, counts the proportion of the number of pixels corresponding to each gray level in the current image information in the image information, sets the initial threshold, calculates the inter-class variance value using a background pixel analysis algorithm, selects the initial threshold corresponding to the maximum inter-class variance value as the first threshold, uses the first threshold to deduce the second threshold according to a preset ratio, compares the gradient amplitude of each initial edge point with the first threshold and the second threshold, and selects the first edge point and the second edge point;

[0008] a parking line recognition unit, wherein after acquiring all the first edge points and the second edge points in the image information, the parking line recognition unit uses a processor to randomly select an edge point, maps the point to the parameter coordinate system in the form of polar coordinates, and counts the number of curves passing through the intersection of the parameter coordinate system. If the number is greater than a preset value, the direction of the straight line in the corresponding plane coordinate system is inferred according to the position of the point, and traverses along the corresponding straight line direction starting from the currently selected edge point. After connecting the edge points in the straight line direction into a line segment, if the spacing between the starting edge point and the ending edge point is greater than a preset length value, the coordinate parameters of the starting edge point and the ending edge point are transmitted to a storage device, and a parking line length range is set. If the difference between the horizontal coordinates of the starting edge point and the ending edge point corresponding to the line segment meets the parking line length range and the vertical coordinate difference is zero, the line segment is determined to be a parking line;

[0009] A parking space coordinate determination unit, wherein the parking space coordinate determination unit sorts all parking lines according to the horizontal coordinate parameters of the starting edge point and the ending edge point, and after the parking lines with the same horizontal coordinates are determined as the same column, each column is numbered one by one, and the minimum vertical coordinate and the maximum vertical coordinate under the same horizontal coordinate are determined, and the corresponding number of parking spaces in each column of the current parking lot is obtained, and the width of the parking space is calculated according to the difference between the minimum vertical coordinate and the maximum vertical coordinate and the number of parking spaces, and the parking line corresponding to each parking space is analyzed according to the parameter coordinates corresponding to the parking line and the parking space width, so as to determine the position coordinates of the parking space and number each parking space.

[0010] Preferably, the edge analysis unit includes a grayscale value calculation module, a grayscale value analysis module and an initial edge point extraction module. After the grayscale value calculation module obtains the collected image information, it sets corresponding weights for the component values ​​corresponding to red, green and blue in the image information, and calculates the grayscale value corresponding to the processed image information according to the component value and weight of each color. The grayscale value analysis module selects a 3×3 filter to sort the grayscale values ​​of each pixel point in the image information and its surrounding adjacent pixels in order from large to small, and after determining the median corresponding to each sequence, the median is used as the grayscale value of the pixel point. The initial edge point extraction module uses a Soble convolution template to calculate the gradient amplitude and direction of each pixel point, and then compares the gradient amplitude of each pixel point with the gradient amplitudes of two pixel points in the same direction. If the gradient amplitude of the point is greater than the gradient amplitudes of the other two points, the point is retained as the initial edge point, otherwise the pixel value of the point is set to zero.

[0011] Preferably, the edge analysis unit further comprises a threshold division module and an edge point determination module, wherein the threshold division module sets a corresponding gray level for each gray value according to the gray value range, and after counting the proportion of the number of pixels corresponding to each gray level in the current image information in the image information, sets an initial threshold, and divides all pixels into two categories according to the initial threshold, namely, non-background pixels and background pixels, and after determining the number of pixels in the two categories and the average gray value, calculates the inter-class variance value according to these data through the background pixel analysis algorithm, and selects the initial threshold corresponding to the largest inter-class variance value. The value is used as the first threshold, and the second threshold is calculated by using the first threshold according to a preset ratio. The edge point determination module compares the gradient amplitude of each initial edge point with the first threshold. If the gradient amplitude of the point is greater than the first threshold, the point is determined to be the first edge point. Otherwise, it continues to be compared with the second threshold. If the point is less than the second threshold, the pixel value of the point is set to zero. Otherwise, it is determined whether the adjacent point is the first edge point. When there is no first edge point in the adjacent point, the pixel value of the point is set to zero. Otherwise, the point is determined to be the second edge point. The background pixel analysis algorithm is specifically as follows:

[0012]

[0013]

[0014] σ 2 =w0w1(u0-u1) 2

[0015] Where w0 represents the proportion of background pixels in the image information, w1 represents the proportion of non-background pixels in the image information, T represents the threshold, i represents the parameter, P(i) represents the proportion of the i-th grayscale level in the image information, N represents the total number of grayscale levels, u0 represents the average grayscale value corresponding to the background pixels, u1 represents the average grayscale value corresponding to the non-background pixels, σ 2 Represents the between-class variance value.

[0016] Preferably, the parking line recognition unit includes a coordinate mapping module and an edge point marking module. After the coordinate mapping module obtains all the first edge points and the second edge points in the image information, it uses a processor to randomly select an edge point to determine whether the point has a mark. If the mark exists, other edge points are reselected. Otherwise, the point is mapped to the parameter coordinate system in the form of polar coordinates. When an intersection appears in the parameter coordinate system, the edge point marking module counts the number of curves passing through the intersection. If the number is greater than a preset value, the direction of the straight line in the corresponding plane coordinate system is calculated according to the position of the point, and traverses along the corresponding straight line direction starting from the currently selected edge point, sets the spacing between edge points, determines all edge points in the straight line direction that meet the spacing, and after connecting these edge points into line segments, marks the starting edge point and the last connected edge point respectively, and deletes all other edge points.

[0017] Preferably, the parking line recognition unit also includes a parameter storage module and a line segment analysis module. If the distance between the starting edge point and the ending edge point is greater than a preset length value, the parameter storage module transmits the coordinate parameters of the starting edge point and the ending edge point to a storage device. After the line segment analysis module determines all the line segments contained in the image information, it sets the parking line length range, and determines whether each line segment is a parking line based on the coordinate parameters of the starting edge point and the ending edge point of each line segment. If the difference between the horizontal coordinates of the starting edge point and the ending edge point corresponding to the line segment meets the parking line length range and the vertical coordinate difference is zero, the line segment is determined to be a parking line. Otherwise, the line segment and the corresponding edge point coordinate parameters are deleted.

[0018] Preferably, the parking space status extraction unit includes an image sample acquisition module and a data set construction module. If the ratio between the number of parked vehicles and the total number of parking spaces is higher than 0.65, the image sample acquisition module uses the OV13850 model camera carried by the drone to shoot the entire parking lot at a preset flight altitude, and transmits the acquired image sample information to a storage device. The data set construction module crops the collected image sample information according to the position coordinates of each parking space, constructs an image data set with vehicles and an image data set without vehicles, and stores the image information of each parking space in the corresponding data set.

[0019] Preferably, the parking space status extraction unit also includes a recognition network generation module and an idle parking space recognition module. The recognition network generation module uses an image data set with vehicles and an image data set without vehicles to train a VGG-16 convolutional neural network to determine the weight parameters in the neural network, thereby obtaining a corresponding parking space image recognition network. After the idle parking space recognition module obtains the image information of the entire area of ​​the current parking lot, it crops the current image information according to the position coordinates of the parking space, analyzes the cropped image information using the parking space image recognition network, determines the image information of the non-existent vehicle, determines the number of the idle parking space based on the image information and position of the non-existent vehicle, and outputs it through a visual interface.

[0020] Preferably, the path analysis unit includes a node setting module, which analyzes the actual coordinates of each parking space according to the location coordinates of the parking space and the flight altitude of the drone when collecting image information, and uses a topological method to build an electronic map of the parking lot according to the actual coordinates of the parking space, and sets multiple nodes in the electronic map. The node types are divided into three types, namely, waiting area nodes, parking space nodes and path nodes.

[0021] Preferably, the path analysis unit also includes a model building module and a vehicle guidance module. The model building module uses a multi-objective optimization algorithm to build a parking space allocation model. The parking space allocation model analyzes the parking path, path length and conflict probability corresponding to each parking path of each vehicle to be parked according to the number of vehicles to be parked, the number of drones and the corresponding nodes of the vacant parking spaces. The vehicle guidance module transmits the parking path and the waiting area node to the drone. The drone guides the vehicle to be parked to the corresponding vacant parking space according to the path information and updates the corresponding parking space information.

[0022] The parking space guidance management method based on drone management includes the following steps:

[0023] S1. Collecting parking area image information: extracting the total number of parking spaces in the parking lot and the number of currently parked vehicles. If the ratio between the number of currently parked vehicles and the total number of parking spaces is less than 0.15, the OV13850 camera carried by the drone is used to shoot the entire parking lot at a preset flight altitude, and the acquired image information is transmitted to a storage device;

[0024] S2. Determine the initial edge point: After acquiring the collected image information, use a 3×3 filter to sort each pixel in the image information and its surrounding adjacent pixels according to the grayscale value, and use the median value of the sequence as the grayscale value of the pixel. Use the Soble convolution template to calculate the gradient amplitude and direction of each pixel to determine the initial edge point;

[0025] S3, screening the first edge point: after counting the proportion of the number of pixels corresponding to each gray level in the current image information in the image information, set the initial threshold, calculate the inter-class variance value through the background pixel analysis algorithm, select the initial threshold corresponding to the largest inter-class variance value as the first threshold, use the first threshold to deduce the second threshold according to a preset ratio, compare the gradient amplitude of each initial edge point with the first threshold and the second threshold, and screen out the first edge point and the second edge point;

[0026] S4, identifying the parking line: after obtaining all the first edge points and the second edge points in the image information, a processor is used to randomly select an edge point, and the point is mapped to the parameter coordinate system in the form of polar coordinates, and the number of curves passing through the intersection of the parameter coordinate system is counted. If the number is greater than a preset value, the direction of the line in the corresponding plane coordinate system is inferred according to the position of the point, and traversal is performed along the corresponding straight line direction starting from the currently selected edge point. After the edge points in the straight line direction are connected into a line segment, if the difference between the horizontal coordinates of the starting edge point and the ending edge point corresponding to the line segment meets the length range of the parking line and the vertical coordinate difference is zero, the line segment is determined to be a parking line;

[0027] S5. Setting parking space numbers: sorting all parking lines according to the horizontal coordinate parameters of the starting edge point and the ending edge point, and after the parking lines with the same horizontal coordinates are determined as the same column, numbering each column one by one, analyzing the parking line corresponding to each parking space according to the parameter coordinates corresponding to the parking line and the parking space width, thereby determining the position coordinates of the parking space, and numbering each parking space;

[0028] S6, identifying vacant parking spaces: after obtaining the image information of the entire area of ​​the current parking lot, cropping the current image information according to the location coordinates of the parking spaces, analyzing the cropped image information using a parking space image recognition network, determining image information where no vehicles exist, determining the number of the vacant parking spaces according to the image information and the location where no vehicles exist, and outputting it through a visual interface;

[0029] S7. Guide the parking path: Analyze the actual coordinates of each parking space according to the location coordinates of the parking space and the flight altitude of the drone when collecting image information. Use the topological method to build an electronic map of the parking lot according to the actual coordinates of the parking space. Analyze the parking path, path length and conflict probability corresponding to each parking path of each vehicle to be parked through the parking space allocation model. Transmit the parking path and waiting area nodes to the drone, and the drone will guide the vehicle to be parked to the corresponding vacant parking space according to the path information.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] 1. The present invention processes the collected image information through an edge analysis unit to screen all edge points in the image information. In order to prevent the influence of noise on image edge detection, the system performs two rounds of edge point screening. In the second screening process, the background analysis algorithm is used to calculate the inter-class variance value according to the image information, and the first threshold and the second threshold corresponding to each image information are further determined. Compared with the method of setting the threshold by itself, the threshold obtained by such a design is more scientific and reasonable, which greatly ensures that the edge points will not be missed during the detection process.

[0032] 2. The present invention utilizes a parking line recognition unit to analyze edge points to obtain all parking lines involved in the image information, and stores the coordinate parameters of each parking line, so that a parking space coordinate determination unit can infer the specific position of the corresponding parking space according to the parking line position and the parking space width. This method of determining parking space coordinates not only saves a lot of manpower, but also can determine the exact position of each parking space in a short time, and does not require the use of too many devices. The path analysis unit obtains the parking path, path length and conflict probability corresponding to each parking path of each vehicle to be parked according to the vacant parking spaces identified by the parking space status extraction unit. On the one hand, it ensures that the drone completes the guidance task, and on the other hand, it predicts the possible congestion probability. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 A schematic diagram of the overall system flow provided by an embodiment of the present invention;

[0034] Figure 2 A block diagram of the internal modules of an edge analysis unit provided by an embodiment of the present invention;

[0035] Figure 3 A block diagram of the internal modules of a parking line recognition unit provided by an embodiment of the present invention;

[0036] Figure 4 A block diagram of the internal modules of a parking space status extraction unit provided in an embodiment of the present invention;

[0037] Figure 5 This is a block diagram of the internal modules of the path analysis unit provided in an embodiment of the present invention.

[0038] In the figure: 1. Image acquisition unit; 2. Edge analysis unit; 201. Gray value calculation module; 202. Gray value analysis module; 203. Initial edge point extraction module; 204. Threshold division module; 205. Edge point determination module; 3. Parking line recognition unit; 301. Coordinate mapping module; 302. Edge point marking module; 303. Parameter storage module; 304. Line segment analysis module; 4. Parking space coordinate determination unit; 5. Parking space status extraction unit; 501. Image sample acquisition module; 502. Data set construction module; 503. Recognition network generation module; 504. Vacant parking space recognition module; 6. Path analysis unit; 601. Node setting module; 602. Model building module; 603. Vehicle guidance module. DETAILED DESCRIPTION

[0039] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0040] See also Figure 1-5 ,The present invention provides a technical solution: a parking space guidance management system based on drone management, including a parking space state extraction unit 5 and a path analysis unit 6;

[0041] Image acquisition unit 1, image acquisition unit 1 extracts the total number of parking spaces in the parking lot and the number of currently parked vehicles. If the ratio between the number of currently parked vehicles and the total number of parking spaces is less than 0.15, the OV13850 camera carried by the drone is used to shoot the entire parking lot at a preset flight altitude, and the acquired image information is transmitted to a storage device;

[0042] Edge analysis unit 2, after acquiring the collected image information, selects a 3×3 filter to sort each pixel in the image information and its surrounding adjacent pixels according to the gray value, and uses the median value corresponding to the sequence as the gray value of the pixel. After calculating the gradient amplitude and direction of each pixel using the Soble convolution template, the initial edge point is determined, and the proportion of the number of pixels corresponding to each gray level in the current image information in the image information is counted, and the initial threshold is set. After calculating the inter-class variance value through the background pixel analysis algorithm, the initial threshold corresponding to the largest inter-class variance value is selected as the first threshold, and the second threshold is calculated according to the first threshold according to the preset ratio. The gradient amplitude of each initial edge point is compared with the first threshold and the second threshold, and the first edge point and the second edge point are screened out;

[0043] The parking line recognition unit 3, after acquiring all the first edge points and the second edge points in the image information, uses the processor to randomly select an edge point, maps the point to the parameter coordinate system in the form of polar coordinates, and counts the number of curves passing through the intersection of the parameter coordinate system. If the number is greater than a preset value, the direction of the straight line in the corresponding plane coordinate system is inferred according to the position of the point, and traverses along the corresponding straight line direction starting from the currently selected edge point. After connecting the edge points in the straight line direction into a line segment, if the spacing between the starting edge point and the ending edge point is greater than a preset length value, the coordinate parameters of the starting edge point and the ending edge point are transmitted to the storage device, and the parking line length range is set. If the difference between the horizontal coordinates of the starting edge point and the ending edge point corresponding to the line segment meets the parking line length range and the vertical coordinate difference is zero, the line segment is determined to be a parking line;

[0044] The parking space coordinate determination unit 4 sorts all parking lines according to the horizontal coordinate parameters of the starting edge point and the ending edge point, and after the parking lines with the same horizontal coordinates are determined as the same column, each column is numbered one by one, and the minimum vertical coordinate and the maximum vertical coordinate under the same horizontal coordinate are determined, and the corresponding number of parking spaces in each column of the current parking lot is obtained, and the width of the parking space is calculated according to the difference between the minimum vertical coordinate and the maximum vertical coordinate and the number of parking spaces, and the parking line corresponding to each parking space is analyzed according to the parameter coordinates corresponding to the parking line and the parking space width, so as to determine the position coordinates of the parking space and number each parking space.

[0045] The edge analysis unit 2 includes a gray value calculation module 201, a gray value analysis module 202 and an initial edge point extraction module 203. After the gray value calculation module 201 obtains the collected image information, it sets corresponding weights for the component values ​​corresponding to red, green and blue in the image information, and calculates the gray value corresponding to the processed image information according to the component value and weight of each color. The gray value analysis module 202 selects a 3×3 filter to sort the gray values ​​of each pixel point in the image information and its surrounding adjacent pixel points in order from large to small, and after determining the median corresponding to each sequence, the median is used as the gray value of the pixel point. The initial edge point extraction module 203 uses the Soble convolution template to calculate the gradient amplitude and direction of each pixel point, and then compares the gradient amplitude of each pixel point with the gradient amplitudes of two pixel points in the same direction. If the gradient amplitude of the point is greater than the gradient amplitudes of the other two points, the point is retained as the initial edge point, otherwise the pixel value of the point is set to zero;

[0046] The edge analysis unit 2 also includes a threshold division module 204 and an edge point determination module 205. The threshold division module 204 sets a corresponding gray level for each gray value according to the gray value range, and after counting the proportion of the number of pixels corresponding to each gray level in the current image information in the image information, sets an initial threshold, and divides all pixels into two categories according to the initial threshold, namely, non-background pixels and background pixels. After determining the number of pixels in the two categories and the average gray value, the background pixel analysis algorithm is used to calculate the inter-class variance value based on these data, and the initial threshold corresponding to the largest inter-class variance value is selected. The initial threshold is used as the first threshold, and the second threshold is calculated by using the first threshold according to a preset ratio. The edge point determination module 205 compares the gradient amplitude of each initial edge point with the first threshold. If the gradient amplitude of the point is greater than the first threshold, the point is determined to be the first edge point. Otherwise, it continues to be compared with the second threshold. If the point is less than the second threshold, the pixel value of the point is set to zero. Otherwise, it is determined whether the adjacent point is the first edge point. When the adjacent point does not have the first edge point, the pixel value of the point is set to zero. Otherwise, the point is determined to be the second edge point. The background pixel analysis algorithm is specifically as follows:

[0047]

[0048]

[0049] σ 2 =w0w1(u0-u1) 2

[0050] Where w0 represents the proportion of background pixels in the image information, w1 represents the proportion of non-background pixels in the image information, T represents the threshold, i represents the parameter, P(i) represents the proportion of the i-th grayscale level in the image information, N represents the total number of grayscale levels, u0 represents the average grayscale value corresponding to the background pixels, u1 represents the average grayscale value corresponding to the non-background pixels, σ 2 Represents the inter-class variance value;

[0051] The parking line recognition unit 3 includes a coordinate mapping module 301 and an edge point marking module 302. After the coordinate mapping module 301 obtains all the first edge points and the second edge points in the image information, it uses a processor to randomly select an edge point to determine whether the point has a mark. If the mark exists, other edge points are reselected. Otherwise, the point is mapped to the parameter coordinate system in the form of polar coordinates. When an intersection appears in the parameter coordinate system, the edge point marking module 302 counts the number of curves passing through the intersection. If the number is greater than a preset value, the direction of the straight line in the corresponding plane coordinate system is calculated according to the position of the point, and traverses along the corresponding straight line direction from the currently selected edge point, sets the spacing between edge points, determines all edge points that meet the spacing in the straight line direction, and after connecting these edge points into line segments, marks the starting edge point and the last connected edge point respectively, and deletes all other edge points.

[0052] The parking line recognition unit 3 also includes a parameter storage module 303 and a line segment analysis module 304. If the distance between the starting edge point and the ending edge point is greater than a preset length value, the parameter storage module 303 transmits the coordinate parameters of the starting edge point and the ending edge point to the storage device. After the line segment analysis module 304 determines all the line segments contained in the image information, it sets the parking line length range, and determines whether each line segment is a parking line according to the coordinate parameters of the starting edge point and the ending edge point of each line segment. If the difference between the horizontal coordinates of the starting edge point and the ending edge point corresponding to the line segment meets the parking line length range and the vertical coordinate difference is zero, the line segment is determined to be a parking line, otherwise the line segment and the corresponding edge point coordinate parameters are deleted;

[0053] The parking space state extraction unit 5 includes an image sample acquisition module 501 and a data set construction module 502. If the ratio between the number of parked vehicles and the total number of parking spaces is higher than 0.65, the image sample acquisition module 501 uses the OV13850 camera carried by the drone to shoot the entire parking lot at a preset flight altitude, and transmits the acquired image sample information to a storage device. The data set construction module 502 crops the acquired image sample information according to the position coordinates of each parking space, constructs an image data set with vehicles and an image data set without vehicles, and stores the image information of each parking space in the corresponding data set;

[0054] The parking space state extraction unit 5 also includes a recognition network generation module 503 and an idle parking space recognition module 504. The recognition network generation module 503 uses the image data set with vehicles and the image data set without vehicles to train the VGG-16 convolutional neural network, determine the weight parameters in the neural network, and thus obtain the corresponding parking space image recognition network. After the idle parking space recognition module 504 obtains the image information of the entire area of ​​the current parking lot, it crops the current image information according to the position coordinates of the parking space, analyzes the cropped image information using the parking space image recognition network, determines the image information of the vehicle without the vehicle, determines the number of the idle parking space according to the image information and the position of the vehicle without the vehicle, and outputs it through a visual interface;

[0055] The path analysis unit 6 includes a node setting module 601. The node setting module 601 analyzes the actual coordinates of each parking space according to the location coordinates of the parking space and the flight altitude of the drone when collecting image information, and uses a topological method to build an electronic map of the parking lot according to the actual coordinates of the parking space, and sets a plurality of nodes in the electronic map. The node types are divided into three types, namely, waiting area nodes, parking space nodes and path nodes.

[0056] The path analysis unit 6 also includes a model building module 602 and a vehicle guidance module 603. The model building module 602 uses a multi-objective optimization algorithm to build a parking space allocation model. The parking space allocation model analyzes the parking path, path length and conflict probability corresponding to each parking path of each vehicle to be parked according to the number of vehicles to be parked, the number of drones and the corresponding nodes of the idle parking spaces. The vehicle guidance module 603 transmits the parking path and the waiting area node to the drone. The drone guides the vehicle to be parked to the corresponding idle parking space according to the path information and updates the corresponding parking space information. The parking space allocation model is specifically as follows:

[0057]

[0058] In the formula, F(r i ) represents the objective function of the parking space allocation model, L(r i ) represents the path length of parking the vehicle to be parked in the vacant parking space, C(r i ) represents the probability of conflict between the path of parking the vehicle to be parked in the vacant parking space and other paths existing in the parking lot at the same time, r i represents the current parking path, i represents the parameter, and n represents the total number of vehicles to be parked;

[0059] The parking space guidance management method based on drone management includes the following steps:

[0060] S1. Collecting parking area image information: extracting the total number of parking spaces in the parking lot and the number of currently parked vehicles. If the ratio between the number of currently parked vehicles and the total number of parking spaces is less than 0.15, the OV13850 camera carried by the drone is used to shoot the entire parking lot at a preset flight altitude, and the acquired image information is transmitted to a storage device;

[0061] S2. Determine the initial edge point: After acquiring the collected image information, use a 3×3 filter to sort each pixel in the image information and its surrounding adjacent pixels according to the grayscale value, and use the median value of the sequence as the grayscale value of the pixel. Use the Soble convolution template to calculate the gradient amplitude and direction of each pixel to determine the initial edge point;

[0062] S3, screening the first edge point: after counting the proportion of the number of pixels corresponding to each gray level in the current image information in the image information, set the initial threshold, calculate the inter-class variance value through the background pixel analysis algorithm, select the initial threshold corresponding to the largest inter-class variance value as the first threshold, use the first threshold to deduce the second threshold according to a preset ratio, compare the gradient amplitude of each initial edge point with the first threshold and the second threshold, and screen out the first edge point and the second edge point;

[0063] S4, identifying the parking line: after obtaining all the first edge points and the second edge points in the image information, a processor is used to randomly select an edge point, and the point is mapped to the parameter coordinate system in the form of polar coordinates, and the number of curves passing through the intersection of the parameter coordinate system is counted. If the number is greater than a preset value, the direction of the line in the corresponding plane coordinate system is inferred according to the position of the point, and traversal is performed along the corresponding straight line direction starting from the currently selected edge point. After the edge points in the straight line direction are connected into a line segment, if the difference between the horizontal coordinates of the starting edge point and the ending edge point corresponding to the line segment meets the length range of the parking line and the vertical coordinate difference is zero, the line segment is determined to be a parking line;

[0064] S5. Setting parking space numbers: sorting all parking lines according to the horizontal coordinate parameters of the starting edge point and the ending edge point, and after the parking lines with the same horizontal coordinates are determined as the same column, numbering each column one by one, analyzing the parking line corresponding to each parking space according to the parameter coordinates corresponding to the parking line and the parking space width, thereby determining the position coordinates of the parking space, and numbering each parking space;

[0065] S6, identifying vacant parking spaces: after obtaining the image information of the entire area of ​​the current parking lot, cropping the current image information according to the location coordinates of the parking spaces, analyzing the cropped image information using a parking space image recognition network, determining image information where no vehicles exist, determining the number of the vacant parking spaces according to the image information and the location where no vehicles exist, and outputting it through a visual interface;

[0066] S7. Guide the parking path: Analyze the actual coordinates of each parking space according to the location coordinates of the parking space and the flight altitude of the drone when collecting image information. Use the topological method to build an electronic map of the parking lot according to the actual coordinates of the parking space. Analyze the parking path, path length and conflict probability corresponding to each parking path of each vehicle to be parked through the parking space allocation model. Transmit the parking path and waiting area nodes to the drone, and the drone will guide the vehicle to be parked to the corresponding vacant parking space according to the path information.

[0067] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0068] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. The parking guidance management system based on drone management is characterized by: It includes a parking space status extraction unit (5) and a path analysis unit (6); An image acquisition unit (1), wherein the image acquisition unit (1) extracts the total number of parking spaces in the parking lot and the number of currently parked vehicles. If the ratio between the number of currently parked vehicles and the total number of parking spaces is less than 0.15, the OV13850 camera carried by the drone is used to photograph the entire parking lot at a preset flight altitude, and the acquired image information is transmitted to a storage device; An edge analysis unit (2), after acquiring the collected image information, the edge analysis unit (2) selects a 3×3 filter to sort each pixel in the image information and its surrounding adjacent pixels according to grayscale values, and uses the median value of the sequence as the grayscale value of the pixel. After calculating the gradient amplitude and direction of each pixel using a Soble convolution template, an initial edge point is determined. After counting the proportion of the number of pixels corresponding to each grayscale level in the current image information in the image information, an initial threshold is set. After calculating the inter-class variance value using a background pixel analysis algorithm, an initial threshold corresponding to the maximum inter-class variance value is selected as a first threshold. The second threshold is calculated using the first threshold according to a preset ratio. The gradient amplitude of each initial edge point is compared with the first threshold and the second threshold to screen out the first edge point and the second edge point. A parking line recognition unit (3), wherein after acquiring all first edge points and second edge points in the image information, the parking line recognition unit (3) randomly selects an edge point using a processor, maps the point to a parameter coordinate system in the form of polar coordinates, counts the number of curves passing through the intersection of the parameter coordinate system, and if the number is greater than a preset value, infers the direction of a straight line in the corresponding plane coordinate system according to the position of the point, traverses along the corresponding straight line direction from the currently selected edge point, connects the edge points in the straight line direction into a line segment, and if the spacing between the starting edge point and the ending edge point is greater than a preset length value, transmits the coordinate parameters of the starting edge point and the ending edge point to a storage device, sets a parking line length range, and if the difference between the horizontal coordinates of the starting edge point and the ending edge point corresponding to the line segment meets the parking line length range and the vertical coordinate difference is zero, determines that the line segment is a parking line; A parking space coordinate determination unit (4), wherein the parking space coordinate determination unit (4) sorts all parking lines according to the horizontal coordinate parameters of the starting edge point and the ending edge point, determines the parking lines with the same horizontal coordinate as the same column, and then numbers each column one by one, determines the minimum vertical coordinate and the maximum vertical coordinate under the same horizontal coordinate, obtains the corresponding number of parking spaces in each column of the current parking lot, calculates the width of the parking space according to the difference between the minimum vertical coordinate and the maximum vertical coordinate and the number of parking spaces, analyzes the parking line corresponding to each parking space according to the parameter coordinates corresponding to the parking line and the parking space width, thereby determining the position coordinates of the parking space, and numbers each parking space.

2. The parking space guidance management system based on drone management according to claim 1 is characterized by: The edge analysis unit (2) comprises a gray value calculation module (201), a gray value analysis module (202) and an initial edge point extraction module (203). After acquiring the collected image information, the gray value calculation module (201) sets corresponding weights for the component values ​​corresponding to red, green and blue in the image information, and calculates the gray value corresponding to the processed image information according to the component value and weight of each color. The gray value analysis module (202) selects a 3×3 filter to sort the gray values ​​of each pixel point in the image information and its surrounding adjacent pixel points in order from large to small, and after determining the median value corresponding to each sequence, the median value is used as the gray value of the pixel point. The initial edge point extraction module (203) calculates the gradient amplitude and direction of each pixel point using a Soble convolution template, and compares the gradient amplitude of each pixel point with the gradient amplitudes of two pixel points in the same direction. If the gradient amplitude of the point is greater than the gradient amplitudes of the other two points, the point is retained as the initial edge point, otherwise the pixel value of the point is set to zero.

3. The parking space guidance management system based on drone management according to claim 2 is characterized in that: The edge analysis unit (2) further comprises a threshold division module (204) and an edge point determination module (205). The threshold division module (204) sets a corresponding grayscale level for each grayscale value according to the grayscale value range, and after counting the proportion of the number of pixels corresponding to each grayscale level in the current image information in the image information, sets an initial threshold, and divides all pixels into two categories, namely non-background pixels and background pixels, according to the initial threshold. After determining the number of pixels in the two categories and the average grayscale value, the background pixel analysis algorithm is used to calculate the inter-class variance value based on these data, and the maximum pixel is selected. The initial threshold value corresponding to the inter-class variance value is used as the first threshold value, and the second threshold value is calculated according to the first threshold value according to the preset ratio. The edge point determination module (205) compares the gradient amplitude of each initial edge point with the first threshold value. If the gradient amplitude of the point is greater than the first threshold value, the point is determined to be a first edge point. Otherwise, it continues to compare with the second threshold value. If the point is less than the second threshold value, the pixel value of the point is set to zero. Otherwise, it is determined whether the adjacent point is the first edge point. When the adjacent point does not have a first edge point, the pixel value of the point is set to zero. Otherwise, the point is determined to be a second edge point.

4. The parking space guidance management system based on drone management according to claim 1 is characterized by: The parking line recognition unit (3) comprises a coordinate mapping module (301) and an edge point marking module (302). After the coordinate mapping module (301) obtains all the first edge points and the second edge points in the image information, it uses a processor to randomly select an edge point and determines whether the point has a mark. If the mark exists, it reselects other edge points. Otherwise, it maps the point to the parameter coordinate system in the form of polar coordinates. When an intersection point appears in the parameter coordinate system, the edge point marking module (302) counts the number of curves passing through the intersection point. If the number is greater than a preset value, it calculates the direction of the straight line in the corresponding plane coordinate system based on the position of the point, starts from the currently selected edge point and traverses along the corresponding straight line direction, sets the spacing between edge points, determines all edge points in the straight line direction that meet the spacing, connects the edge points into line segments, respectively marks the starting edge point and the last connected edge point, and deletes the other edge points.

5. The parking space guidance management system based on drone management according to claim 4 is characterized in that: The parking line recognition unit (3) further comprises a parameter storage module (303) and a line segment analysis module (304). If the distance between the starting edge point and the ending edge point is greater than a preset length value, the parameter storage module (303) transmits the coordinate parameters of the starting edge point and the ending edge point to a storage device. After determining all the line segments contained in the image information, the line segment analysis module (304) sets a parking line length range and determines whether each line segment is a parking line based on the coordinate parameters of the starting edge point and the ending edge point of each line segment. If the difference between the horizontal coordinates of the starting edge point and the ending edge point corresponding to the line segment meets the parking line length range and the vertical coordinate difference is zero, the line segment is determined to be a parking line. Otherwise, the line segment and the corresponding edge point coordinate parameters are deleted.

6. The parking space guidance management system based on drone management according to claim 1 is characterized by: The parking space state extraction unit (5) comprises an image sample acquisition module (501) and a data set construction module (502). If the ratio between the number of parked vehicles and the total number of parking spaces is higher than 0.65, the image sample acquisition module (501) uses the OV13850 camera carried by the drone to shoot the entire parking lot at a preset flight altitude, and transmits the acquired image sample information to a storage device. The data set construction module (502) crops the acquired image sample information according to the position coordinates of each parking space, constructs an image data set with vehicles and an image data set without vehicles, and stores the image information of each parking space in the corresponding data set.

7. The parking space guidance management system based on drone management according to claim 6 is characterized by: The parking space state extraction unit (5) further comprises a recognition network generation module (503) and an idle parking space recognition module (504). The recognition network generation module (503) trains a VGG-16 convolutional neural network using an image data set with a vehicle present and an image data set without a vehicle present, determines weight parameters in the neural network, and thereby obtains a corresponding parking space image recognition network. The idle parking space recognition module (504) obtains image information of the entire area of ​​the current parking lot, crops the current image information according to the position coordinates of the parking space, analyzes the cropped image information using the parking space image recognition network, determines image information of the vehicle not present, determines the number of the idle parking space according to the image information and the position of the vehicle not present, and outputs it through a visual interface.

8. The parking space guidance management system based on drone management according to claim 1 is characterized by: The path analysis unit (6) comprises a node setting module (601), wherein the node setting module (601) analyzes the actual coordinates of each parking space according to the position coordinates of the parking space and the flight altitude of the drone when collecting image information, constructs an electronic map of the parking lot according to the actual coordinates of the parking space using a topological method, and sets a plurality of nodes in the electronic map, wherein the node types are divided into three types, namely, a waiting area node, a parking space node and a path node.

9. The parking space guidance management system based on drone management according to claim 8 is characterized in that: The path analysis unit (6) also includes a model building module (602) and a vehicle guidance module (603). The model building module (602) uses a multi-objective optimization algorithm to build a parking space allocation model. The parking space allocation model analyzes the parking path, path length and conflict probability corresponding to each parking path of each vehicle to be parked according to the number of vehicles to be parked, the number of drones and the corresponding nodes of the idle parking spaces. The vehicle guidance module (603) transmits the parking path and the waiting area node to the drone. The drone guides the vehicle to be parked to the corresponding idle parking space according to the path information and updates the corresponding parking space information.

10. The parking space guidance management method based on drone management according to any one of claims 1 to 9, characterized in that The following steps are involved: S1. Collecting parking area image information: extracting the total number of parking spaces in the parking lot and the number of currently parked vehicles. If the ratio between the number of currently parked vehicles and the total number of parking spaces is less than 0.15, the OV13850 camera carried by the drone is used to shoot the entire parking lot at a preset flight altitude, and the acquired image information is transmitted to a storage device; S2. Determine the initial edge point: After acquiring the collected image information, use a 3×3 filter to sort each pixel in the image information and its surrounding adjacent pixels according to the grayscale value, and use the median value of the sequence as the grayscale value of the pixel. Use the Soble convolution template to calculate the gradient amplitude and direction of each pixel to determine the initial edge point; S3, screening the first edge point: after counting the proportion of the number of pixels corresponding to each gray level in the current image information in the image information, set the initial threshold, calculate the inter-class variance value through the background pixel analysis algorithm, select the initial threshold corresponding to the largest inter-class variance value as the first threshold, use the first threshold to deduce the second threshold according to a preset ratio, compare the gradient amplitude of each initial edge point with the first threshold and the second threshold, and screen out the first edge point and the second edge point; S4, identifying the parking line: after obtaining all the first edge points and the second edge points in the image information, a processor is used to randomly select an edge point, and the point is mapped to the parameter coordinate system in the form of polar coordinates, and the number of curves passing through the intersection of the parameter coordinate system is counted. If the number is greater than a preset value, the direction of the line in the corresponding plane coordinate system is inferred according to the position of the point, and traversal is performed along the corresponding straight line direction starting from the currently selected edge point. After the edge points in the straight line direction are connected into a line segment, if the difference between the horizontal coordinates of the starting edge point and the ending edge point corresponding to the line segment meets the length range of the parking line and the vertical coordinate difference is zero, the line segment is determined to be a parking line; S5. Setting parking space numbers: sorting all parking lines according to the horizontal coordinate parameters of the starting edge point and the ending edge point, and after the parking lines with the same horizontal coordinates are determined as the same column, numbering each column one by one, analyzing the parking line corresponding to each parking space according to the parameter coordinates corresponding to the parking line and the parking space width, thereby determining the position coordinates of the parking space, and numbering each parking space; S6, identifying vacant parking spaces: after obtaining the image information of the entire area of ​​the current parking lot, cropping the current image information according to the location coordinates of the parking spaces, analyzing the cropped image information using a parking space image recognition network, determining image information where no vehicles exist, determining the number of the vacant parking spaces according to the image information and the location where no vehicles exist, and outputting it through a visual interface; S7. Guide the parking path: Analyze the actual coordinates of each parking space according to the location coordinates of the parking space and the flight altitude of the drone when collecting image information. Use the topological method to build an electronic map of the parking lot according to the actual coordinates of the parking space. Analyze the parking path, path length and conflict probability corresponding to each parking path of each vehicle to be parked through the parking space allocation model. Transmit the parking path and waiting area nodes to the drone, and the drone will guide the vehicle to be parked to the corresponding vacant parking space according to the path information.

Images