Garage intelligent parking guiding method based on alteration principle and optimal parking path planning algorithm

Abstract

The invention discloses a garage intelligent parking guidance method based on an alteration principle and an optimal parking path planning algorithm, which is characterized in that the method adopts a method of fusing multi-modal information such as graphics, vision and voice to realize intelligent parking guidance of a garage, and specifically comprises the following steps: 1) constructing a garage electronic map, and providing visual available parking space information for a vehicle owner; 2) realizing intelligent license plate recognition and real-time driving trajectory tracking by utilizing a visual detection and analysis algorithm; 3) according to an alteration principle and real-time vehicle positioning, path searching is carried out in a half space of a garage plane by adopting shortest parking path planning of a Dijkstra algorithm, and efficient parking navigation of a multi-layer garage is supported at the same time; compared with the prior art, the method has the advantages of being clear in available parking position, intelligent in path planning, convenient and fast in navigation prompting and the like, an effective solution is provided for the parking difficulty of the underground garage, and the method is helpful for improving the living satisfaction of people.

Description

Technical Field

[0001] The present invention relates to the technical field of visual perception and digital image processing, and in particular to a license plate character recognition algorithm based on a support vector machine and a garage intelligent parking guidance method based on the altruistic principle and an optimal parking path planning algorithm. Background Art

[0002] With the development of society and economy and the improvement of people's living standards, cars have become one of the main ways for people to travel. The surge in the number of cars (especially small passenger cars) has led to an increasingly prominent contradiction between too many cars and too few parking spaces, which is more obvious in first-tier cities where every inch of land is valuable. Underground garages are common parking places in people's daily lives. When users park in underground garages, they usually face three difficulties: 1) The number of parking spaces is limited, which makes it difficult to meet parking needs, and vehicles cannot even enter the garage, which is mainly due to the limitation of physical space capacity. 2) The communication quality of underground garages is poor, and the commonly used GPS positioning is easily obstructed by buildings, which reduces the positioning accuracy, which increases the difficulty of real-time and accurate navigation. 3) When parking, the driver cannot obtain the available parking location information and arrival path in time, and mainly relies on searching for available parking spaces within the field of vision. Because underground garages have poor lighting conditions, complex building structures and parking space distribution, and many vehicles entering and leaving, this method of finding vacant parking spaces is time-consuming and laborious, which can easily lead to drivers getting lost, congested passages, and driving safety issues.

[0003] At present, there are some guidance measures for parking in underground garages, such as indicating the availability of parking spaces by red and green lights above them. There have been some related studies at home and abroad, including: 1) In terms of parking systems, the introduction of parking guidance systems can display parking space information and reasonably disperse vehicle flow. The new parking system based on RFID positioning uses image processing technology to guide vehicles through data recorded by cameras. 2) In terms of positioning technology, in addition to the above-mentioned GPS positioning, there are also infrared positioning, ultrasonic positioning, Bluetooth positioning, RFID positioning and WiFi positioning. By comparing these methods, it can be found that WiFi indoor positioning has the advantages of many terminals, high accuracy, not easily affected by buildings, and low cost. It can meet daily needs and can be applied to the current parking guidance system. 3) License plate recognition is the most critical part of the intelligent transportation system and a typical application of computer vision technology, including the detection of numbers and texts, as well as important steps such as semantic recognition. The license plate recognition method based on the improved AlexNet can avoid the influence of license plate character segmentation and complex background environment on the recognition results. The recognition accuracy rate is as high as 98.03%, which has high reliability. For large-angle license plate recognition, researchers proposed an improved Faster RCNN and YOLO deep learning license plate recognition algorithm, which increased the accuracy of license plate recognition at large angles to 99.7%, and adopted a lightweight framework for easy application deployment. 4) Path planning is a strategy to connect the sequence points between the starting position and the end position under the constraints of a specific application environment. Mainstream path planning algorithms include A* algorithm, Dijkstra algorithm, PRM algorithm, neural network algorithm, ant colony algorithm, etc.

[0004] However, when car owners enter the garage, they still face difficulties such as unclear available parking spaces, unintelligent route planning, and inconvenient navigation prompts.

[0005] In summary, it is necessary to develop and optimize the parking guidance technology of underground garages based on artificial intelligence while making full use of existing space resources, reasonably allocate parking resources for users, and automatically plan the shortest driving route so that they can reach available parking spaces as quickly and safely as possible. Through the parking guidance method and system of the garage, the utilization efficiency of the parking lot can be improved, the cost of human management can be reduced, and it has important practical significance and application value for the improvement of social development and people's life satisfaction. Summary of the invention

[0006] The purpose of the present invention is to provide a garage intelligent parking guidance method based on the altruistic principle and the optimal parking path planning algorithm to address the deficiencies of the prior art. The method adopts computer vision license plate recognition and tracking algorithms to construct a garage intelligent parking guidance system with a mobile user end and a server end. The user binds the mobile phone and the vehicle through the mobile terminal to obtain real-time path guidance information. The server end completes the parking space management based on the electronic map of the garage, reasonably allocates parking resources for the user, and automatically plans the shortest driving route so that the user can reach the available parking space as quickly and safely as possible. The method is easy to use and has strong practicality. It can improve the utilization efficiency of the parking lot and reduce the cost of human management. It has important practical significance and application value for social development and the improvement of people's life satisfaction.

[0007] The purpose of the present invention is achieved as follows: a garage intelligent parking guidance method based on the altruism principle and the optimal parking path planning algorithm, which is characterized by adopting a method of multimodal information fusion such as graphics, vision, and voice to build a garage intelligent parking guidance system with a mobile user end and a server end. The user realizes the binding of the mobile phone and the vehicle through the mobile terminal to obtain real-time path guidance information; the server end completes the parking space management based on the garage electronic map, the license plate recognition and tracking algorithm based on computer vision, and the optimal parking path planning algorithm. The mobile user end completes the binding of the mobile terminal and the driven vehicle by scanning the parking code. The garage intelligent parking guidance method based on the altruism principle and visual perception technology specifically includes the following steps:

[0008] Step 1: Scan the parking code to complete the binding between the mobile terminal and the vehicle being driven;

[0009] Step 2: According to the electronic map of the garage, use the parking space allocation strategy based on the altruism principle to allocate available parking space resources for the vehicles in step 1;

[0010] Step 3: Based on the parking space allocated in step 2, the backend server calls the parking path planning algorithm to calculate the navigation path, and displays the obtained optimal parking path information on the large screen above the entrance;

[0011] Step 4: Use the parking path planning algorithm to calculate the best driving route, and update the driving route based on the current position of the vehicle collected in real time while maintaining the stability of the route;

[0012] Step 5: The backend server calls the license plate recognition algorithm based on computer vision and support vector machine, uses machine learning methods to identify the license plate number, and at the same time, tracks the movement of the vehicle according to the collection equipment at each intersection to provide the current location of the vehicle for step 4;

[0013] Step 6: Based on the best driving route calculated in step 4, voice navigation is performed on the mobile terminal to realize intelligent parking guidance in the garage.

[0014] The parking space allocation strategy based on the altruism principle is that the system will give priority to allocating available parking spaces in the distance to the vehicles that arrive first, and guide the vehicle to drive to the allocated parking space. When there are multi-story parking spaces, the available parking spaces in the distance refer to the parking spaces in the lower-level parking spaces. During driving, users can independently choose available parking spaces in addition to allocated parking spaces, and their parking information is also sensed and recorded in the parking space management module of the server.

[0015] The parking path information includes: the current position, the target parking space position, and whether it is necessary to pass through the vehicle passage between two floors.

[0016] The parking path planning algorithm calculates the best driving path according to the target parking space position and the current position of the vehicle, specifically including:

[0017] 1) Determine whether it is necessary to pass through the interlayer channel C. If necessary, first drive from the current position to the interlayer channel Cs of the current layer, and then reach the interlayer channel Ce of the target layer;

[0018] 2) On each layer, the Dijkstra algorithm is run to calculate the shortest distance from the current position of the vehicle to the target position. When the Dijkstra algorithm is executed, the garage space is divided into two parts according to the longitudinal line of the entrance gate, and the side S with a larger total number of available parking spaces is selected as the vehicle passage space.

[0019] The license plate recognition algorithm based on computer vision and support vector machine includes: a license plate position detection algorithm and a license plate character recognition algorithm. The license plate position detection algorithm uses a method based on license plate color and size repositioning combined with a recognition algorithm in the open source computer vision library OpenCV; the license plate prediction algorithm uses a support vector machine SVM to realize the recognition of Chinese characters, letters and numbers in the license plate.

[0020] The voice navigation is to obtain the optimal path through structured processing to form a sentence such as: x meters ahead, turn left or right, update the variables in the sentence in real time according to the planned path, and finally convert the obtained sentence into voice and broadcast it to the user through the mobile user terminal.

[0021] Compared with the prior art, the present invention has the following beneficial technical effects and significant technical progress:

[0022] 1) Introducing advanced artificial intelligence technology into garage parking, realizing intelligent guided parking based on visual perception, accurately recognizing license plates and locating vehicles, updating path planning algorithms based on visual positioning information, and providing intuitive visual driving path information, such as large-screen visual information guidance. These technologies help realize unmanned and intelligent garage management.

[0023] 2) The parking space allocation and path planning algorithm based on the altruistic principle allocates available parking spaces in the distance to vehicles that arrive early, provides more convenient parking solutions for vehicles that arrive later, saves their parking time, and supports dynamic and smooth updates of the path based on real-time perception results.

[0024] 3) The combination of voice reminders and visual observation ensures driving safety. According to the driving route planning and current position, the safe and convenient guidance is achieved by using the coordinated use of vision and voice. Path reminders are given through voice broadcasts without distracting the driver's attention.

[0025] 4) It solves the problem that users cannot intuitively find available parking locations, especially in multi-storey garage spaces. There is a lack of humanized and intelligent path planning algorithms when users move from their current location to the planned parking location. In particular, when driving in a garage, the navigation prompts are not convenient and the safety is not high. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 A schematic diagram of the structure of the intelligent parking guidance system for a garage according to the present invention;

[0027] Figure 2 This is the flow chart of the license plate recognition algorithm;

[0028] Figure 3 Scan the code to obtain the location map;

[0029] Figure 4 Take a photo of the license plate to identify the image;

[0030] Figure 5 Start navigation interface diagram for large screen display and mini program;

[0031] Figure 6 This is a schematic diagram of the B1 path planning for the underground parking garage;

[0032] Figure 7 This is a schematic diagram of the B2 path planning for the underground parking garage;

[0033] Figure 8 This is a schematic diagram of a vehicle arriving at a target parking space. DETAILED DESCRIPTION

[0034] In order to facilitate the understanding of the present invention, the present invention is described in detail below with reference to the accompanying drawings and embodiments.

[0035] Example 1

[0036] See also Figure 1Referring to the whole process of cars entering underground garages in daily life, the present invention designs a garage intelligent parking guidance method framework with a mobile user end and a server end. Users can bind their mobile phones and vehicles through mobile terminals to obtain real-time path guidance information; the server end completes parking space management based on the electronic map of the garage, license plate recognition and tracking algorithms based on computer vision, and optimal parking path planning algorithms. This framework is suitable for parking guidance in multi-story garage structures.

[0037] 1. Mobile User Terminal

[0038] 1-1: Scan the parking code to complete the binding between the mobile terminal and the vehicle being driven

[0039] At the entrance stage, there is usually a gate, and the camera detects the license plate number of the vehicle entering the garage. At this time, by scanning the parking code, the mobile app is started, and only the customer's mobile phone number is obtained, which is convenient for subsequent sending of navigation instructions and conversion into voice prompts. In this way, the mobile phone and the vehicle are bound to each other during the entry stage ( Figure 1 The arrow in the figure indicates 1).

[0040] 1-2: Driving route prompts

[0041] According to the camera collection equipment at the main intersection, the license plate number is collected in real time to relocate the vehicle ( Figure 1 The middle arrow indicates 5). Then, the path planning algorithm is used to refresh the driving path ( Figure 1 At this time, the mobile terminal will display the real-time status and driving path of the vehicle through graphics ( Figure 1 The arrow in the figure indicates 7).

[0042] 1-3: Voice navigation

[0043] After calculating the new driving path, the mobile terminal can broadcast navigation information ( Figure 1 The arrow mark 8 in the figure is similar to the navigation map app commonly used by drivers, but with simplified functions. Voice navigation is suitable for underground garages. The combination of voice and user's line of sight enables multi-channel information transmission, which can reduce the degree to which users rely on vision to find parking spaces.

[0044] 2. Server side

[0045] 2-1: Parking space management

[0046] According to the spatial structure of the garage, the system constructs an electronic map, marking the garage's lanes, the location of each parking space and its corresponding parking space number. In actual operation, the system marks the occupancy status of each parking space (free, occupied, pending). When a vehicle enters, the available parking space resources are allocated or recommended according to the principle of altruism ( Figure 1 The arrow in the figure indicates 2).

[0047] 2-2: Parking path planning algorithm

[0048] Based on the available parking spaces and the current location of the vehicle, the parking path planning algorithm is used to calculate the optimal driving route ( Figure 1 As the vehicle is constantly moving, the system can refresh the route based on the current position of the vehicle collected in real time while maintaining a stable route. At this point, the system can display the parking route on a large screen. Based on the parking space management information in the background, the background server arranges available parking spaces and calculates the navigation path, which is displayed on a large-size display above the driver's head at the entrance ( Figure 1 The arrow mark 4). Urban viaducts are generally equipped with real-time traffic electronic display screens, which display information such as road direction and real-time traffic conditions. The large screen display of this system has similar functions to the display screen on the viaduct, providing drivers with guidance on the overall direction of travel.

[0049] 2-3: License Plate Recognition and Tracking Algorithm

[0050] The license plate recognition and tracking algorithm is based on the principle of computer vision and uses machine learning methods to identify license plate numbers. At the same time, the system tracks the movement of vehicles based on the collection equipment at each intersection and provides vehicle positioning ( Figure 1 The arrow in the figure indicates 5). The license plate recognition and tracking algorithm specifically includes:

[0051] 2-3-1: License plate recognition algorithm based on computer vision and support vector machine

[0052] See also Figure 2 The license plate recognition algorithm based on computer vision and support vector machine includes two parts: license plate position detection algorithm and license plate character recognition algorithm. In the license plate position detection algorithm, a method based on license plate color and size relocation is used, which is combined with the traditional recognition algorithm in the open source computer vision library OpenCV. The license plate prediction algorithm uses support vector machine SVM to realize the recognition of Chinese characters, letters and numbers in the license plate.

[0053] 2-3-2: License plate position detection algorithm

[0054] In a garage environment, we are generally faced with situations such as low light intensity, inconsistent camera angles for capturing vehicles, and different resolutions of camera images. At this time, if we only rely on the traditional OpenCV algorithm for edge detection, the positioning effect achieved by the system may not be good. Therefore, the present invention proposes a method for re-positioning based on license plate color and size. In this scheme, blue, green and yellow license plates are identified, and the read RGB image is converted into the HSV color mode, where (R, G, B) are the normalized components of the red, green and blue channels of the image, respectively. This method mainly further screens the located area by setting the upper and lower limits of the H (hue) component to exclude areas that are not red, green or blue license plates.

[0055] 2-3-3: License Plate Character Recognition Algorithm

[0056] In view of the uneven angles of license plate shooting, the recognition system needs to correct the located license plate to make it in the horizontal direction. Then, the system implements the license plate character segmentation through the steps of graying, binarization, projection analysis, threshold segmentation, etc. for the located license plate area. Since the intervals between the characters of the current standard license plate are not the same, there is a spacer between the second and third characters, so it is necessary to perform an open operation to eliminate the influence of the spacer during character segmentation. The license plate characters are mainly segmented by the histogram and projection analysis of the grayscale image. First, the image is projected horizontally to determine the upper and lower borders of the characters. Then, the image is projected vertically and each character is segmented according to the threshold determined by the histogram. Then, the segmented characters are sent to the trained SVM classifier for character recognition. In order to distinguish easily confused characters, such as the character "2" and the character "Z", the character "8" and the character "B", etc., the present invention adopts secondary classification to further improve the accuracy of character recognition. In addition, license plate recognition has regional characteristics. The license plates of various countries are composed of different characters. my country's license plates are composed of Chinese characters, letters and numbers. The main problems in license plate recognition are as follows: due to the complexity of Chinese characters, high character similarity, fuzzy loss of license plates and other reasons, the difficulty of license plate recognition increases. In order to solve the problem of misrecognition of similar characters in license plate images, the present invention uses ST loss to train the SVM classifier. ST loss essentially maximizes the probability and sum of all paths. Its characteristics are the introduction of blank characters and the ability to achieve rapid calculation of gradients through recursive operations. ST loss draws on the forward and backward algorithm ideas (Forward-Backward) of the Hidden Markov Model HMM (Hidden Markov Model), and uses a dynamic programming algorithm to calculate the ST Loss function and its derivatives, so that the SVM classifier can accurately identify similar characters. Finally, in order to reduce the occurrence of overfitting during training and make the training results more robust, the present invention introduces a weight decay term in the loss function, that is, adds an L2 norm regularization term on the basis of the original loss function.

[0057] Assuming that the output condition of the SVM classifier at each moment is independent for a given input sequence and model index, the probability function can be defined as the output probability of a single character by the following formula:

[0058]

[0059] in, : Input data, such as a set of character images of a license plate; is the license plate character sequence, whose length is . ; For the character position, the SVM predicts that the character belongs to the category of the entire license plate sequence probability is defined by the following formula:

[0060]

[0061] where is the complete license plate output (such as "Lu B1**45"). is all possible license plate set of character sequences.

[0062] Assuming that the output conditions of the SVM classifier at each moment are independent given the input sequence and the model index, the probability function can be defined by the following formula:

[0063] .

[0064] ST Loss ensures that the probability of the finally output license plate is maximized by maximizing and its loss function form is expressed by the following formula:

[0065] .

[0066] where is the training set.

[0067] Expand the above formula by logarithmic operation and express it by the following formula:

[0068] .

[0069] Substitute 's expression into the above formula, and the following formula can be obtained:

[0070]

[0071] Substitute the single-character output probability into ST Loss, and the following formula for L(S) can be obtained:

[0072] .

[0073] where the path probability is the product of all character classification probabilities.

[0074] According to the above relationship, the following formula for L(S) can be obtained:

[0075] .

[0076] 2-4: Optimal parking path planning algorithm based on the altruism principle

[0077] When parking in a garage, the starting point (i.e., source point) is determined. Parking path planning needs to focus on two aspects: determining the target parking space (i.e., end point) and designing the shortest path. The following is a detailed analysis of each aspect.

[0078] Generally speaking, when a car user sees a parking space, he will park nearby. Although this solves the problem of individual parking, it causes the remaining available parking spaces to be irregularly distributed. Moreover, compared with the users who arrive first, it is more difficult for the users who arrive later to search for a parking space in a more limited and disordered set of available parking spaces. At the same time, the users who arrive later are in a relatively more anxious (or nervous) mood when looking for parking spaces. Inspired by this, the present invention designs a parking space allocation strategy based on the principle of altruism, that is, the system allocates available parking spaces in the distance to the vehicles that arrive first, and guides the vehicle to drive to the allocated parking space. Of course, during the driving process, the user can choose an available parking space to park independently in addition to the allocated parking space, and his parking information is also perceived and recorded in the parking space management module of the server segment. There are many mature algorithms for the shortest path design. The present invention selects the classic Dijkstra algorithm, which has high algorithm efficiency and is used to calculate the single-source shortest path in a weighted directed graph. Therefore, the present invention adopts the improved Dijkstra algorithm for parking path planning. The main steps of the algorithm are as follows:

[0079] 1) Call the parking space management module to determine the farthest available parking space Pi (marked as free at this time) as the end point of the path navigation.

[0080] 2) According to the position of Pi, determine whether the position (set as being located at the Bi layer) and the current position S (set as being located at the L1 layer) are on the same layer. If they are not on the same layer, obtain the position of a vehicle channel (C) between the two layers, and determine that the positions of the L1 layer and the Bi layer in channel C are C0 and Ci respectively. The navigation route is divided into three sections: (P1) driving from S to C0, (P2) then C0 drives along the channel to Ci, and (P3) finally driving from Ci to Pi. Both P1 and P3 go to step (3) to execute the path planning algorithm.

[0081] 3) Divide the garage space into two parts according to the longitudinal line of the entrance gate; select the side S with the largest number of available parking spaces as the vehicle passage space.

[0082] 4) In space S, run the Dijkstra algorithm to solve the shortest driving route and determine the shortest path planning from the source point to the destination.

[0083] 5) Set Pi's flag to pending.

[0084] 6) When the vehicle reaches Pi, parking guidance ends and the mark of Pi is set to occupied.

[0085] 2-5: Voice broadcast navigation information

[0086] The optimal path obtained by the above algorithm needs to be broadcast to the user in the form of voice, so the navigation information needs to be structured. First, a fixed sentence pattern is imported, such as: x meters ahead, turn (left / right). Then, according to the planned path, the variables in the sentence pattern are updated in real time. Finally, the obtained sentence is converted into voice and broadcast to the user through the mobile application.

[0087] Example 2

[0088] 1. Design of small programs for mobile users

[0089] See also Figure 3 The applet is designed based on the wxml framework and JavaScript script. After opening the applet, enter the main page. According to the page prompts, the user scans the parking code to obtain the user's location permission (bound to the mobile phone number) and current location.

[0090] See also Figure 4 After obtaining the user's location, the license plate recognition link is entered. This step will be completed by the garage hardware and the server. In order to simulate the entire process, the present invention also designs this function on the user side. Then, the user clicks "Photo Identification" to display the license plate photo captured by the camera, and the system begins to plan parking spaces and navigation routes. At this time, the user can click the "Navigation Simulation" button to display the navigation route. Due to possible external interference such as weak GPS signals, the system uses a registration-free wireless signal connection. When entering the parking lot, the user can complete the operation of connecting to the WiFi signal.

[0091] (II) Large screen display

[0092] See also Figure 5 a. After the user's license plate is recognized and the position is located, the "parking space management" module selects an available parking space based on the principle of altruism and allocates it to the current vehicle. Here it is shown that the current available parking space is B2, so you need to drive to the entrance first.

[0093] See also Figure 5 b. Drive to position A009 on the B2 layer. Then, the user clicks "Start Navigation" to start driving navigation.

[0094] 3. License Plate Recognition

[0095] The present invention is implemented based on Python programming environment and OpenCV software library. The program performs model training and license plate detection on the device (CPU: i7-9750H@2.6 GHz, memory: 16 GB, operating system: Windows 10). The license plate recognition algorithm uses 7286 license plate images in the basic subset of CCPD 2019 for training and detection, of which 5464 are used as model training sets and the remaining 1822 are used as test sets. The test images are divided into two categories, 1500 images with better quality are selected as the first category, and 322 images with relatively poor quality are selected as the second category (the license plate has a certain degree of blur, and the license plate inclination is in the range of -2~6 degrees). Through testing, the scheme has a recognition accuracy rate of more than 95% for license plates. By statistically analyzing the recognition rates of different types of characters, the results show that due to the complexity of Chinese characters, their correct recognition rate is slightly lower than that of English characters and numbers. The experimental results show that when the license plate image has a heavy blur and there are many similar characters, the algorithm used in the present invention has a better recognition effect. For example, the character "8", the character "S", and the character "B" are similar characters that are easily misrecognized.

[0096] 4. Path planning

[0097] In order to determine the vehicle's driving path, the system needs to abstract the space of each floor of the garage into a plane model, and then find the optimal path based on the optimal parking path planning algorithm.

[0098] See also Figure 6-7 ,The system assumes that the underground parking garage has two floors (B1 and B2) and two entrances and exits, and there is a traffic channel (C) between the two floors. According to the optimal parking path planning algorithm based on the altruistic principle, this system needs to guide the first-arriving vehicles to the available parking spaces farther away.

[0099] See also Figure 6 ,Take the parking space on the B2 floor of the parking garage as an example, the system first calculates the shortest distance from the entrance to the inter-layer channel C by running the Dijkstra algorithm.

[0100] See also Figure 7 , since the farthest upper right corner parking space on the B2 floor is already occupied by the red car, the system will allocate the parking space P1 on the left side of the red car as the target parking spot. Therefore, the system runs the Dijkstra algorithm again to guide the vehicle from C to P1. In view of the above garage environment settings, the present invention compares the performance of the traditional Dijkstra algorithm and the Dijkstra algorithm based on the altruism principle (using binary space division). The test indicators are respectively at B1 and B2, planning the entrance to Figure 5The driving route of parking space P1 in the figure is calculated, and the time taken is calculated. The improved Dijkstra algorithm has a great improvement in running time, which is less than 50% of the traditional Dijkstra algorithm, basically achieving the expected goal.

[0101] 5. Voice navigation

[0102] During navigation, the route map is displayed on the user's mobile phone screen, and navigation is performed through voice broadcast. First, a fixed sentence pattern is imported, such as: x meters ahead, turn (left / right). Then, according to the planned path, the variables in the sentence pattern are updated in real time. Finally, the obtained sentence is converted into voice and broadcast to the user through the mobile application.

[0103] (VI) Parking ends

[0104] See also Figure 8 , after arriving at the parking space (parking next to the red car in the upper right corner), click the Done button to end navigation.

[0105] The above is only a preferred implementation of the present invention and does not limit the present invention in any form. Therefore, any simple modification, equivalent change and modification made to the above implementation method according to the technical essence of the present invention without departing from the content of the technical solution of the present invention shall be included in the scope of the claims of the present invention as long as it is an equivalent implementation of the present invention.

Claims

1. A garage intelligent parking guidance method based on the altruistic principle and the optimal parking path planning algorithm, characterized in that: The method adopts a multimodal information fusion method of graphics, vision and voice to realize intelligent parking guidance in the garage, which specifically includes the following steps: Step 1: Scan the parking code to complete the binding between the mobile terminal and the vehicle being driven; Step 2: According to the electronic map of the garage, use the parking space allocation strategy based on the altruism principle to allocate available parking space resources for the vehicles in step 1; Step 3: Based on the parking space allocated in step 2, the backend server calls the parking path planning algorithm to calculate the navigation path, and displays the obtained optimal parking path information on the large screen above the entrance; Step 4: Use the parking path planning algorithm to calculate the best driving route, and update the driving route based on the current position of the vehicle collected in real time while maintaining the stability of the route; Step 5: The backend server calls the license plate recognition algorithm based on computer vision and support vector machine, uses machine learning methods to identify the license plate number, and at the same time, tracks the movement of the vehicle according to the collection equipment at each intersection to provide the current location of the vehicle for step 4; Step 6: Based on the best driving route calculated in step 4, voice navigation is performed on the mobile terminal to realize intelligent parking guidance in the garage.

2. The intelligent parking guidance method for a garage based on the altruistic principle and the optimal parking path planning algorithm according to claim 1 is characterized in that: The parking space allocation strategy based on the altruism principle is that the system will give priority to allocating available parking spaces in the distance to the vehicles that arrive first, and guide the vehicle to drive to the allocated parking space. When there are multi-story parking spaces, the available parking spaces in the distance refer to the parking spaces in the lower-level parking spaces. During driving, users can independently choose available parking spaces in addition to allocated parking spaces, and their parking information is also sensed and recorded in the parking space management module of the server.

3. The intelligent parking guidance method for a garage based on the altruistic principle and the optimal parking path planning algorithm according to claim 1, characterized in that: The parking path information includes: the current position, the target parking space position, and whether it is necessary to pass through the vehicle passage between two floors.

4. The intelligent parking guidance method for a garage based on the altruistic principle and the optimal parking path planning algorithm according to claim 1, characterized in that: The parking path planning algorithm calculates the best driving path according to the target parking space position and the current position of the vehicle, specifically including: 1) Determine whether it is necessary to pass through the interlayer channel C. If necessary, first drive from the current position to the interlayer channel Cs of the current layer, and then reach the interlayer channel Ce of the target layer; 2) On each layer, the Dijkstra algorithm is run to calculate the shortest distance from the current position of the vehicle to the target position. When the Dijkstra algorithm is executed, the garage space is divided into two parts according to the longitudinal line of the entrance gate, and the side S with a larger total number of available parking spaces is selected as the vehicle passage space.

5. The intelligent parking guidance method for a garage based on the altruistic principle and the optimal parking path planning algorithm according to claim 1, characterized in that: The license plate recognition algorithm based on computer vision and support vector machine includes: a license plate position detection algorithm and a license plate character recognition algorithm. The license plate position detection algorithm uses a method based on license plate color and size repositioning combined with a recognition algorithm in the open source computer vision library OpenCV; the license plate prediction algorithm uses a support vector machine SVM to realize the recognition of Chinese characters, letters and numbers in the license plate.

6. The intelligent parking guidance method for a garage based on the altruistic principle and the optimal parking path planning algorithm according to claim 1, characterized in that: The voice navigation is to obtain the optimal path through structured processing to form a sentence such as: x meters ahead, turn left or right, update the variables in the sentence in real time according to the planned path, and finally convert the obtained sentence into voice and broadcast it to the user through the mobile user terminal.

Images