New energy automobile automatic parking method based on multistage feature fusion and intelligent optimization

Abstract

The invention discloses a new energy automobile automatic parking method based on multi-level feature fusion and intelligent optimization, an intelligent optimization system and an automatic parking system based on multi-level feature fusion realize automatic control of an automobile parking strategy, and the method comprises the following steps of environment sensing, decision making and planning, and control execution and optimization and adjustment. According to the method, a feature fusion mode of a sensor system is combined, system optimization is carried out on the basis, more accurate and comprehensive environment features are obtained, vehicle body postures and dynamic obstacle data serve as influence conditions of parking track dynamic planning, real-time parking track dynamic adjustment is achieved, more efficient path planning is achieved, and the parking track dynamic planning efficiency is improved. And meanwhile, the robustness and the safety of the parking system are also improved.

Description

Technical Field

[0001] The present invention relates to the technical field of automatic parking of automobiles, and in particular to an automatic parking method for new energy automobiles based on multi-level feature fusion and intelligent optimization. Background Art

[0002] With the continuous development of driverless technology, the application prospects of automatic parking technology in the market are becoming more and more broad. Automatic parking technology can improve parking efficiency, reduce the waste of parking space, and solve the problem of parking difficulties. For certain driving groups, driverless automatic parking technology is a convenient solution. In addition, driverless automatic parking technology will also play an important role in the field of shared travel, improving service quality and user experience.

[0003] Automatic parking technology aims to enable vehicles to automatically identify parking spaces and complete parking operations. Currently, it faces some challenges, such as: the low accuracy of parking space detection and identification, and the shortcomings of ultrasonic sensors such as large beam angle, poor directionality, low resolution, and short range. For example, when detecting oblique or narrow parking spaces, inaccurate detection may occur due to beam angle and resolution issues; there is also the problem of insufficient adaptability to complex scenes. When encountering some more complex situations, such as oblique parking spaces, too narrow parking spaces, or no obstacles around the parking spaces that can be located, automatic parking may fail due to inability to identify, and may increase the risk of accidents and reduce the driving experience.

[0004] In this way, automatic parking technology is needed to solve these problems in the future development, and the automatic parking control system based on multi-information fusion is an intelligent system developed to solve a series of problems in the automatic parking process. With the development of automobile driving towards intelligence, the automatic parking system has received widespread attention, but there are currently problems such as low parking accuracy and inability to effectively identify irregular parking spaces. Multi-information fusion technology provides ideas for solving these problems. Summary of the invention

[0005] 1. Technical issues to be solved

[0006] In view of the deficiencies in the prior art, the present invention adopts an intelligent optimization system with multi-level feature fusion and an automatic parking system to realize automatic control of vehicle parking strategies, further improve the accuracy and safety of automatic parking, and bring safety protection to the automatic parking system.

[0007] (II) Technical solution

[0008] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: a method for automatic parking of new energy vehicles based on multi-level feature fusion and intelligent optimization, an intelligent optimization system based on multi-level feature fusion and an automatic parking system to realize automatic control of vehicle parking strategy, comprising the following steps;

[0009] 1) Perception of the environment

[0010] The sensor system distributed around the vehicle detects the surrounding environment information in real time, such as obstacles, parking lines and other vehicles, measures the distance between the car and surrounding objects, and analyzes the images returned by the visual camera to convert the ground parking lines into a three-dimensional model.

[0011] 2) Decision-making and planning

[0012] The central processor analyzes and processes the collected data to obtain the current position of the car, the target position and the surrounding environmental parameters. The control system makes decisions and plans according to the preset algorithm, determines the size and position of the parking space and the position of obstacles, and calculates the optimal parking path;

[0013] 3) Control execution

[0014] According to the above parameters, the automatic parking strategy is made and converted into electrical signals. After receiving the electrical signals, the vehicle strategy control system executes the steering, acceleration and braking actions of the vehicle according to the decision and planning results of the control system to realize automatic parking;

[0015] 4) Optimization and adjustment

[0016] During the automatic parking process, the vehicle body posture is perceived by combining and complementing the data information from multiple sensors. Based on the multi-sensor data fusion, the feature-level fusion algorithm is used to comprehensively analyze the features extracted by different sensors to obtain more accurate and comprehensive environmental features. The vehicle body posture and dynamic obstacle data are used as influencing conditions for the dynamic planning of the parking trajectory, realizing real-time dynamic adjustment of the parking trajectory.

[0017] As a preferred solution, the process of establishing the three-dimensional model in the perception environment is as follows:

[0018] A1. Use mapping to locate the vehicle and scan the parking map;

[0019] A2. Collect surrounding environment information using ultrasonic sensors, cameras and body sensors;

[0020] A3. Scan and identify available parking spaces, while ensuring real-time positioning and tracking of the vehicle itself, and completing the perception and modeling of information around the vehicle.

[0021] As a preferred solution, in decision-making and planning, a parking strategy is used for automatic parking, and the process is as follows:

[0022] B1. Use vehicle sensors to perceive the vehicle's surrounding environment and establish a world perception information fusion model;

[0023] B2. Use vehicle-mounted ultrasonic radar and visual sensors, combined with multi-dimensional data fusion analysis and intelligent parking space recognition algorithms to achieve perception of the automatic parking environment around the vehicle;

[0024] B3. Identify relevant characteristic information of the parking space surrounding environment, and calculate and output the parking space recognition result based on parking data analysis.

[0025] As a preferred solution, the control system is based on visual technology and artificial intelligence algorithms, which will evaluate whether the current environment is suitable for parking and select a suitable parking space. It also includes the execution and control of the electric power steering system, electronic stability program, automatic transmission and braking system.

[0026] As a preferred solution, the perceived world information fusion model performs statistics and analysis on the perception results, including:

[0027] Visual perception, obtaining road information, traffic signs and other vehicle information through cameras;

[0028] Sound perception, used in vehicle anti-collision alarms and voice recognition;

[0029] LiDAR sensing, which uses laser beams to measure distance and obtain three-dimensional information about the parking environment;

[0030] Millimeter-wave radar perception is used for target detection and tracking in parking environments.

[0031] As a preferred solution, the intelligent optimization system is based on information fusion technology and includes the following steps:

[0032] C1. Use multiple sensors to model the parking space model and extract characteristic parameters, mathematically model typical irregularly parked vehicles on both sides of the target parking space, and output the vehicle teaching model;

[0033] C2. Through geometric analysis, we found various factors that affect parking space recognition: vehicle body posture angle, maximum width distance, shortest space distance and the presence or absence of parking space lines;

[0034] C3. Use the environmental data perceived by visual and ultrasonic sensors to identify parking spaces of various shapes, colors and sizes, including but not limited to white, yellow, parallel, vertical, diagonal and special parking spaces. Use deep learning methods to train parking space analysis and dynamic obstacle analysis models in a targeted manner.

[0035] C3, and through automatic obstacle avoidance algorithm and high-precision vehicle posture positioning, the vehicle's automatic parking operation trajectory is optimized and adjusted.

[0036] As a preferred solution, the establishment of the teaching model requires modeling the parking space model and extracting characteristic parameters. These characteristic parameters include but are not limited to the vehicle body posture angle, maximum width distance, shortest space distance and whether there is a parking space line. A coordinate system is used to represent the position and posture of the vehicle, and geometric shapes (such as rectangles, ellipses, etc.) are used to approximate the contour of the vehicle.

[0037] As a preferred solution, the cubic spline interpolation method is used in C3 to smooth the parking path to ensure smooth changes in direction when the vehicle is driving. Given a series of (x i ,y i ); cubic spline interpolation constructs a piecewise cubic polynomial function s i (x), so that every interval [x i , x i+1 The inner s(x) is a cubic polynomial and satisfies the boundary conditions and continuity conditions. The specific formula is as follows:

[0038] s i (x) = a i +b i (xx i )+c i (xx i )+d i (xx i )

[0039] where a i , b i , c i and d i are unknown coefficients, which are solved by boundary conditions and continuity conditions.

[0040] (III) Beneficial effects

[0041] Compared with the prior art, the present invention provides a new energy vehicle automatic parking method based on multi-level feature fusion and intelligent optimization, which has the following beneficial effects:

[0042] 1. The present invention adopts a multi-level feature fusion approach, utilizes vehicle-mounted ultrasonic radar and visual sensors, combines multi-dimensional data fusion analysis and intelligent parking space recognition algorithms, realizes the perception of the automatic parking environment around the vehicle, provides a more comprehensive environmental perception capability, further improves parking accuracy and enhances adaptability to complex environments.

[0043] 2. The present invention combines the feature fusion method of the sensor system and optimizes the system on this basis to obtain more accurate and comprehensive environmental characteristics, and uses the vehicle body posture and dynamic obstacle data as influencing conditions for dynamic planning of parking trajectory, thereby realizing real-time dynamic adjustment of parking trajectory, achieving more efficient path planning, and also improving the robustness and safety of the parking system. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a schematic diagram of the process flow of the automatic parking method of the present invention;

[0045] Figure 2 Schematic diagram showing the functions of the automatic parking system of the present invention. DETAILED DESCRIPTION

[0046] In order to better understand the purpose, structure and function of the present invention, the following will further illustrate the automatic parking method for new energy vehicles based on multi-level feature fusion and intelligent optimization of the present invention in combination with the accompanying drawings and specific embodiments.

[0047] Example 1

[0048] refer to Figure 1-2 , the present invention provides a new energy vehicle automatic parking method based on multi-level feature fusion and intelligent optimization, an intelligent optimization system based on multi-level feature fusion and an automatic parking system to realize automatic control of vehicle parking strategy, including the following steps;

[0049] 1) Perception of the environment

[0050] Through the sensor system distributed around the vehicle, it detects the surrounding environment information in real time, such as obstacles, parking lines and other vehicles, measures the distance between the car and surrounding objects, and performs image analysis on the images returned by the visual camera to convert the ground parking lines into a three-dimensional model.

[0051] The process of establishing the three-dimensional model is as follows:

[0052] A1. Use mapping to locate the vehicle and scan the parking map;

[0053] A2. Collect surrounding environment information using ultrasonic sensors, cameras and body sensors;

[0054] A3. Scan and identify available parking spaces, while ensuring real-time positioning and tracking of the vehicle itself, and completing the perception and modeling of information around the vehicle.

[0055] The core of mapping and positioning lies in simultaneous localization and mapping (SLAM). This technology provides high-precision positioning and mapping services by fully utilizing the advantages of each sensor through multi-sensor fusion.

[0056] 2) Decision-making and planning

[0057] The central processing unit analyzes and processes the collected data to derive the car's current position, target position, and surrounding environmental parameters. The control system makes decisions and plans based on preset algorithms, determines the size and position of the parking space and the location of obstacles, and calculates the optimal parking path.

[0058] Among them, the parking strategy is used for automatic parking, and the process is as follows:

[0059] B1. Use vehicle sensors to perceive the vehicle's surrounding environment and establish a world perception information fusion model;

[0060] B2. Use vehicle-mounted ultrasonic radar and visual sensors, combined with multi-dimensional data fusion analysis and intelligent parking space recognition algorithms to achieve perception of the automatic parking environment around the vehicle;

[0061] B3. Identify relevant characteristic information of the parking space surrounding environment, and calculate and output the parking space recognition result based on parking data analysis.

[0062] Among them, the perception world information fusion model is to count and analyze the perception results, including:

[0063] Visual perception, obtaining road information, traffic signs and other vehicle information through cameras;

[0064] Sound perception, used in vehicle anti-collision alarms and voice recognition;

[0065] LiDAR sensing, which uses laser beams to measure distance and obtain three-dimensional information about the parking environment;

[0066] Millimeter-wave radar perception is used for target detection and tracking in parking environments.

[0067] 3) Control execution

[0068] An automatic parking strategy is made based on the above parameters and converted into electrical signals. After receiving the electrical signals, the vehicle strategy control system executes the steering, acceleration, braking and other actions of the vehicle according to the decision and planning results of the control system to achieve automatic parking.

[0069] 4) Optimization and adjustment

[0070] During the automatic parking process, the vehicle body posture is perceived by combining and complementing data information from multiple sensors. Based on multi-sensor data fusion, the feature-level fusion algorithm conducts a comprehensive analysis of the features extracted by different sensors to obtain more accurate and comprehensive environmental features. The vehicle body posture and dynamic obstacle data are used as influencing conditions for the dynamic planning of the parking trajectory, realizing real-time dynamic adjustment of the parking trajectory.

[0071] Specifically, the control system of the present invention is based on visual technology and artificial intelligence algorithms, and will evaluate whether the current environment is suitable for parking and select a suitable parking space, and also includes the execution and control of the electric power steering system, electronic stability program, automatic transmission and braking system. This includes sensors such as ultrasonic radar, millimeter wave radar, laser radar and camera around the vehicle, which detect the surrounding environment in real time, and make decisions and plans through the control system. Finally, the execution subsystem controls the steering, acceleration and braking of the vehicle to achieve automatic parking.

[0072] In the present invention, environmental perception is a crucial link. Through sensors such as ultrasonic radar, the system can detect obstacles around the vehicle body in real time to avoid collisions. At the same time, these sensors can also detect parking spaces before parking begins, and provide necessary lateral obstacle information during parking to ensure the safety and accuracy of parking. Specifically, ultrasonic radar is used to measure short distances, the camera identifies the parking line and converts it into a three-dimensional model, and the parking space position and angle are calculated based on the data from various sensors. Then the collected data is handed over to the central processor. After obtaining environmental information, the control system of the automatic parking system will make decisions and plans according to the preset algorithm. The control system will evaluate whether the current environment is suitable for parking, select a suitable parking space, and plan the best path from the current position to the parking space. In this process, the control system needs to fully consider factors such as the vehicle's dynamic characteristics, safety, and parking lot rules.

[0073] Example 2

[0074] like Figure 2 As shown, the intelligent optimization system of the present invention is based on information fusion technology and includes the following steps:

[0075] C1. Use multiple sensors to model the parking space model and extract characteristic parameters, mathematically model typical irregularly parked vehicles on both sides of the target parking space, and output the vehicle teaching model;

[0076] C2. Through geometric analysis, we found various factors that affect parking space recognition: vehicle body posture angle, maximum width distance, shortest space distance and the presence or absence of parking space lines;

[0077] C3. Use the environmental data perceived by visual and ultrasonic sensors to identify parking spaces of various shapes, colors and sizes, including but not limited to white, yellow, parallel, vertical, diagonal and special parking spaces. Use deep learning methods to train parking space analysis and dynamic obstacle analysis models in a targeted manner.

[0078] C3, and through automatic obstacle avoidance algorithm and high-precision vehicle posture positioning, the vehicle's automatic parking operation trajectory is optimized and adjusted.

[0079] In this embodiment, its intelligent optimization system adopts a multi-level feature fusion approach, that is, it collects data information from multiple sensors of the vehicle for complementation and combination to obtain a more accurate description of the environment, and combines the body camera to perceive the surrounding environment, and combines multi-dimensional data fusion analysis and intelligent parking space recognition and detection algorithms to achieve perception of the automatic parking environment around the vehicle, and achieves high-precision vehicle motion control through automatic obstacle avoidance algorithms and high-precision body posture positioning.

[0080] The specific process is as follows: an information fusion of ultrasonic sensors, visual information sensors and vehicle odometers is input into the system, and then the system performs fuzzy reasoning. The system will determine the type of parking space and then plan the path for automatic parking of the vehicle. If the planning is successful, the perception ends. Otherwise, it continues to move forward to detect the parking space until the detection is completed.

[0081] More specifically, the establishment of the teaching model of this embodiment requires modeling the parking space model and extracting characteristic parameters, which include but are not limited to the body posture angle, maximum width distance, shortest space distance and whether there is a parking space line. The coordinate system is used to represent the position and posture of the vehicle, and geometric shapes (such as rectangles, ellipses, etc.) are used to approximate the outline of the vehicle. The established parking space model can more completely describe the parking space scene, and use the fuzzy reasoning method that imitates the thinking characteristics of the driver to perform multi-sensor information fusion. The proposed intelligent automatic parking space recognition method can effectively identify regular and irregular parking space types.

[0082] Furthermore, in this embodiment C3, the cubic spline interpolation method is used to smooth the parking path to ensure that the direction of the vehicle changes smoothly when driving. Given a series of (x i ,y i ); cubic spline interpolation constructs a piecewise cubic polynomial function s i (x), so that every interval [x i , x i+1 The inner s(x) is a cubic polynomial and satisfies the boundary conditions and continuity conditions. The specific formula is as follows:

[0083] s i (x) = a i +b i (xx i )+c i (xx i )+d i (xx i )

[0084] where a i , b i , c i and di are unknown coefficients, which are solved by boundary conditions and continuity conditions.

[0085] The cubic spline interpolation algorithm is an interpolation method that approximates a given set of data points by constructing a series of cubic polynomials, and ensures that the interpolation function value at the data point is equal to the data value, while maintaining the continuity of the second-order derivative over the entire interval. In this embodiment, the algorithm improves the accuracy and smoothness of parking by generating a smooth and continuous parking trajectory.

[0086] It is to be understood that the present invention is described by some embodiments, and it is known to those skilled in the art that various changes or equivalent substitutions may be made to these features and embodiments without departing from the spirit and scope of the present invention. In addition, under the teachings of the present invention, these features and embodiments may be modified to adapt to specific circumstances and materials without departing from the spirit and scope of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the scope of protection of the present invention.

Claims

1. A new energy vehicle automatic parking method based on multi-level feature fusion and intelligent optimization, an intelligent optimization system based on multi-level feature fusion and an automatic parking system to realize automatic control of vehicle parking strategy, characterized in that: The following steps are involved: 1) Perception of the environment The sensor system distributed around the vehicle detects the surrounding environment information in real time, such as obstacles, parking lines and other vehicles, measures the distance between the car and surrounding objects, and analyzes the images returned by the visual camera to convert the ground parking lines into a three-dimensional model. 2) Decision-making and planning The central processor analyzes and processes the collected data to obtain the current position of the car, the target position and the surrounding environmental parameters. The control system makes decisions and plans according to the preset algorithm, determines the size and position of the parking space and the position of obstacles, and calculates the optimal parking path; 3) Control execution According to the above parameters, the automatic parking strategy is made and converted into electrical signals. After receiving the electrical signals, the vehicle strategy control system executes the steering, acceleration and braking actions of the vehicle according to the decision and planning results of the control system to realize automatic parking; 4) Optimization and adjustment During the automatic parking process, the vehicle body posture is perceived by combining and complementing the data information from multiple sensors. Based on the multi-sensor data fusion, the feature-level fusion algorithm is used to comprehensively analyze the features extracted by different sensors to obtain more accurate and comprehensive environmental features. The vehicle body posture and dynamic obstacle data are used as influencing conditions for the dynamic planning of the parking trajectory, realizing real-time dynamic adjustment of the parking trajectory.

2. The automatic parking method for new energy vehicles based on multi-level feature fusion and intelligent optimization according to claim 1 is characterized in that: The process of establishing the three-dimensional model in the perception environment is as follows: A1. Use mapping to locate the vehicle and scan the parking map; A2. Collect surrounding environment information using ultrasonic sensors, cameras and body sensors; A3. Scan and identify available parking spaces, while ensuring real-time positioning and tracking of the vehicle itself, and completing the perception and modeling of information around the vehicle.

3. The automatic parking method for new energy vehicles based on multi-level feature fusion and intelligent optimization according to claim 2 is characterized in that: In decision-making and planning, parking strategies are used for automatic parking, and the process is as follows: B1. Use vehicle sensors to perceive the vehicle's surrounding environment and establish a world perception information fusion model; B2. Use vehicle-mounted ultrasonic radar and visual sensors, combined with multi-dimensional data fusion analysis and intelligent parking space recognition algorithms to achieve perception of the automatic parking environment around the vehicle; B3. Identify relevant characteristic information of the parking space surrounding environment, and calculate and output the parking space recognition result based on parking data analysis.

4. The automatic parking method for new energy vehicles based on multi-level feature fusion and intelligent optimization according to claim 3 is characterized in that: The control system is based on visual technology and artificial intelligence algorithms. It will evaluate whether the current environment is suitable for parking and select a suitable parking space. It also includes the execution and control of the electric power steering system, electronic stability program, automatic transmission and braking system.

5. The automatic parking method for new energy vehicles based on multi-level feature fusion and intelligent optimization according to claim 3 is characterized in that: The perceived world information fusion model performs statistics and analysis on the perceived results, including: Visual perception, obtaining road information, traffic signs and other vehicle information through cameras; Sound perception, used in vehicle anti-collision alarms and voice recognition; LiDAR sensing, which uses laser beams to measure distance and obtain three-dimensional information about the parking environment; Millimeter-wave radar perception is used for target detection and tracking in parking environments.

6. The automatic parking method for new energy vehicles based on multi-level feature fusion and intelligent optimization according to claim 1 is characterized in that: The intelligent optimization system is based on information fusion technology and includes the following steps: C1. Use multiple sensors to model the parking space model and extract characteristic parameters, mathematically model typical irregularly parked vehicles on both sides of the target parking space, and output the vehicle teaching model; C2. Through geometric analysis, we found various factors that affect parking space recognition: vehicle body posture angle, maximum width distance, shortest space distance and the presence or absence of parking space lines; C3. Use the environmental data perceived by visual and ultrasonic sensors to identify parking spaces of various shapes, colors and sizes, including but not limited to white, yellow, parallel, vertical, diagonal and special parking spaces. Use deep learning methods to train parking space analysis and dynamic obstacle analysis models in a targeted manner. C3, and through automatic obstacle avoidance algorithm and high-precision vehicle posture positioning, the vehicle's automatic parking operation trajectory is optimized and adjusted.

7. The automatic parking method for new energy vehicles based on multi-level feature fusion and intelligent optimization according to claim 6 is characterized in that: The establishment of the teaching model requires modeling the parking space model and extracting characteristic parameters, which include but are not limited to the vehicle body posture angle, maximum width distance, shortest space distance and whether there is a parking space line. A coordinate system is used to represent the position and posture of the vehicle, and geometric shapes (such as rectangles, ellipses, etc.) are used to approximate the contour of the vehicle.

8. The automatic parking method for new energy vehicles based on multi-level feature fusion and intelligent optimization according to claim 6 is characterized in that: The cubic spline interpolation method is used in C3 to smooth the parking path and ensure smooth changes in the direction of the vehicle when driving. Given a series of (x i ,y i ); Cubic spline interpolation constructs a piecewise cubic polynomial function s i (x), so that every interval [x i , x i+1 ] s(x) is a cubic polynomial and satisfies the boundary conditions and continuity conditions. The specific formula is as follows: s i (x)=a i +b i (x-x i )+c i (x-x i )+d i (x-x i ) where a i , b i , c i and d i are unknown coefficients, which are solved by boundary conditions and continuity conditions.

Images