Automatic parking method based on self-selected parking space

Abstract

The present invention discloses an automatic parking method based on self-selected parking spaces, which formally eliminates the process of searching for parking spaces globally in the parking lot and weakens the requirement for real-time parking space identification. While reducing the system's operational pressure, it can design more complex functions for assisted parking. Specifically, an inner point is obtained in the desired parking space selected by the user, a first region of interest is delineated based on the inner point, and a second region of interest containing the inner point is identified. The second region of interest that meets the confidence condition is released to the user, or the confidence threshold is lowered to detect the second region of interest again; after the user accepts and selects the released second region of interest, parking path planning is performed based on the second region of interest. The concept of the present invention is to shift from global parking space identification to target parking space detection in the area surrounding a fixed point within the self-selected parking space, thereby reducing a large number of features involved in parking space identification, thereby being able to identify parking spaces that are more difficult to identify.

Description

Technical Field

[0001] The present invention relates to the technical field of assisted parking, and in particular to an automatic parking method based on self-selected parking spaces. Background Art

[0002] As the number of cars increases, the market demand for parking assistance functions is also increasing. This paper studies and analyzes several existing mainstream automatic parking assistance solutions as follows:

[0003] (1) Automatic parking assistance systems based on ultrasonic radar or 360-degree panoramic bird's-eye view. This technical solution is characterized by using a perception system composed of radar or cameras in conjunction with an algorithm that can identify parking spaces to search for parking spaces near the car while driving at low speeds. The identified parking space information is then passed to the planning and control modules to complete the entire parking process. This is a commonly used technical solution. Its disadvantage is that it has high requirements for the perception module and parking space recognition algorithm, and is usually limited to identifying only standard parking spaces.

[0004] (2) Based on the existing automatic parking assistance, a memory parking solution is implemented by combining the construction and memory capabilities of local maps. This solution aims to solve the problem of locating a specific parking space and automatically parking in indoor parking lots with weak GPS signals. When implementing this solution, a complete parking process record is required, and it needs to be supported by a powerful perception system and computing unit. It is usually combined with cloud services and intelligent transportation to provide better positioning and parking effects.

[0005] (3) Autonomous valet parking system. This technical solution relies on the vehicle’s own sensors, parking lot positioning signs, parking lot high-precision maps and cloud scheduling to jointly realize the autonomous valet parking function. It is a highly intelligent and highly networked and mutually synergistic concept solution. It is difficult to promote and popularize it at this stage.

[0006] (4) The self-selected parking space intelligent parking control system mainly uses the method of pasting a map on a multimedia device to select a parking space for the vehicle to be parked directly. This solution requires the driver to manually determine the position of the car after parking. However, on the one hand, the driver needs to consider the position, direction and obstacles of the parking space when pasting the rectangular map, which will inevitably increase the threshold of use; on the other hand, it is difficult for the driver to stand in the system during the process of determining the parking space. Considering that the trajectory planning and posture control in the actual parking process are both difficult, the actual application of this solution often results in forced interruption due to the inability to complete the planning.

[0007] (5) The automatic parking space display system based on panoramic images mainly uses panoramic images to search for parking spaces in real time and display them. This solution requires the driver to drive at a low speed in order to find a parking space. In order for the system to have a good auxiliary parking effect, the requirements for the false detection rate and system real-time performance are both high. Moreover, in an environment with a large number of standard vacant parking spaces, the driver will continuously receive parking space information, which brings troubles and obstacles to targeted parking space identification.

[0008] (6) Automatic parking system based on parking space corner detection. This solution mainly builds a feature detector to obtain a heat map of the physical features of the parking space, such as corners, entrance lines, and side lines, and then determines the parking area based on the information in the heat map. The idea of ​​using corner information combined with entrance lines and side lines to determine parking spaces is a conventional method for parking space recognition based on visual schemes. However, the factors that determine the recognition success rate are not only the quality of the algorithm itself, but also depend on the properties of the target parking space. That is, this solution is only effective for standard parking spaces where corners can be detected.

[0009] After comprehensively analyzing the above automatic parking assistance solutions, the present invention summarizes the pain points of mainstream technical solutions in the parking space search process:

[0010] On the one hand, the existing parking space search process mostly uses a global perspective to search, which not only requires a large amount of feature data but is also more suitable for parking spaces that are easier to identify;

[0011] On the other hand, when searching for global parking spaces, available parking spaces are constantly released. Most of this information is not what the driver needs, and the parking spaces that the driver needs may not be accurately released and provided to the driver for final selection due to various factors.

[0012] On the other hand, in order to ensure the real-time release of available parking spaces and give drivers time and space to park and choose a parking space, the existing parking space search process not only increases system pressure, especially the units responsible for perception and calculation, but also requires the vehicle to maintain a very low speed, otherwise it is easy to miss the released parking space. Summary of the Invention

[0013] In view of the above, the present invention aims to provide an automatic parking method based on self-selected parking space to solve the above-mentioned technical problems.

[0014] The technical solution adopted in the present invention is as follows:

[0015] The present invention provides an automatic parking method based on self-selected parking space, which includes:

[0016] After the self-selected parking space function of the parking system is activated, receiving the inner point of the desired parking space selected by the user;

[0017] Taking the inner point as the geometric center, a rectangular range is set around it to obtain a first region of interest;

[0018] Invoking a preset parking space recognition algorithm to detect a second region of interest for parking within the first region of interest; wherein the second region of interest includes the interior point;

[0019] If, after this round of detection, the second region of interest output by the parking space recognition algorithm meets the current confidence threshold, it is released to the user; the confidence threshold is used to verify the confidence of the output result of the parking space recognition algorithm;

[0020] If the second ROI output by the parking space recognition algorithm is not released to the user after this round of detection, lowering the confidence threshold and detecting the second ROI again from within the first ROI;

[0021] If the parking space recognition algorithm does not output the second region of interest after this round of detection, the user is prompted to reselect a desired parking space;

[0022] After the user accepts and selects a released second region of interest, a predicted parking position and direction within the selected second region of interest are determined based on the vehicle's front orientation and the posture type of the selected second region of interest, and parking path planning is performed based on the predicted parking position and direction.

[0023] In at least one possible implementation manner, when the confidence threshold is reduced to a preset critical value, if the second region of interest has not been released, a pending parking area replacing the second region of interest is allocated to the user.

[0024] In at least one possible implementation, a method for obtaining the pending parking area includes:

[0025] Searching for a plurality of edge lines adjacent to the interior point in the first region of interest by using a color gradient change and / or edge detection and / or line extraction algorithm;

[0026] The found edge line is used to delineate an area having the same size and shape as the preset standard parking space.

[0027] In at least one possible implementation manner, a size of the rectangular range is greater than or equal to a size of two preset standard parking spaces.

[0028] In at least one possible implementation manner, the posture types include vertical, horizontal, and tilted.

[0029] In at least one possible implementation manner, if the second region of interest output by the parking space recognition algorithm meets a current confidence threshold, releasing it to the user includes: preferentially releasing the second region of interest of a vertical type.

[0030] In at least one possible implementation manner, lowering the confidence threshold includes lowering the confidence threshold according to a preset gradient.

[0031] The main design concept of the present invention is that it formally eliminates the process of searching for parking spaces globally in the parking lot, and also weakens the requirement for real-time parking space identification. While reducing the operating pressure of the system, the saved computing power can be used to design more complex functions for assisted parking. Specifically, an inner point is obtained in the desired parking space selected by the user, a first region of interest is delineated based on the inner point, and a second region of interest containing the inner point is identified. The second region of interest that meets the confidence condition is released to the user, or the confidence threshold is lowered and the second region of interest is detected again; after the user accepts and selects the released second region of interest, parking path planning is performed based on the second region of interest. The present invention changes the underlying idea of ​​traditional parking space identification, from global parking space identification to target parking space detection in the area surrounding a fixed point in the self-selected parking space, thereby reducing a large number of features involved in parking space identification, thereby being able to identify parking spaces that are more difficult to identify. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be further described below with reference to the accompanying drawings, in which:

[0033] Figure 1 This is a flow chart of an automatic parking method based on self-selected parking space provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0034] The following describes embodiments of the present invention in detail. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended only to explain the present invention and are not to be construed as limiting the present invention.

[0035] Before the present invention is elaborated, it can be explained that those skilled in the art will understand that the following embodiments of the present invention must be implemented in combination with hardware carriers. For example, (1) a perception module, such as fisheye cameras with a viewing angle close to or equal to 180° equipped on the front, left, rear, and right sides of the vehicle, and a radar perception system, which may refer to a laser radar, a millimeter-wave radar deployed on the periphery of the vehicle, or an ultrasonic radar; (2) a display module, such as a display with a touch screen function and a display host capable of processing touch signals and video signals; (3) a parking control module, such as a central computing platform that can be used for vehicle control or a domain controller dedicated to the parking system, or other electronic control units for controlling the vehicle.

[0036] The fisheye camera of the aforementioned perception module can transmit video signals to the parking control module. These video signals can generate 360-degree panoramic images, which are used in combination with radar to complete parking space recognition and obstacle recognition. In addition to processing the information of the perception signals, the parking control module is also responsible for decision-making during the parking process, path planning and vehicle posture control. The display module is mainly used for display and completing human-computer interaction.

[0037] Therefore, based on but not limited to the above hardware system architecture, the present invention proposes an embodiment of an automatic parking method based on self-selected parking space. Specifically, Figure 1 shown, including:

[0038] Step S1: After the self-selected parking space function of the parking system is activated, the user's desired parking space point is received;

[0039] Step S2: taking the inner point as the geometric center and setting a rectangular range around it to obtain a first region of interest;

[0040] Step S3: calling a preset parking space recognition algorithm to detect a second region of interest for parking within the first region of interest; wherein the second region of interest includes the interior point;

[0041] Step S41: If, after this round of detection, the second region of interest output by the parking space recognition algorithm meets the current confidence threshold, the second region of interest is released to the user; the confidence threshold is used to verify the confidence of the output result of the parking space recognition algorithm;

[0042] Step S42: if the second ROI output by the parking space recognition algorithm is not released to the user after this round of detection, lower the confidence threshold and detect the second ROI again from within the first ROI;

[0043] Step S43: If the parking space recognition algorithm does not output the second region of interest, prompt the user to reselect a desired parking space;

[0044] Step S5: After the user accepts and locks the second ROI, a predicted parking position and direction within the second ROI is determined based on the vehicle's frontal orientation and the type of the second ROI, and point-to-point parking motion planning is performed based on the predicted parking position and direction.

[0045] To facilitate understanding of the aforementioned embodiments, the present invention provides the following specific description for implementation reference:

[0046] The driver can use the parking system or the vehicle's display module to interact with the user and enter the parking interface. This interface displays a real-time panoramic image and a live model of the surrounding environment provided by radar sensing (this is prior art and not the focus of this invention). The driver then selects the "Self-Select Parking Space" option to activate the self-select parking feature (for safety reasons, this process may be subject to certain restrictions, such as safety prompts such as "Please Apply Brakes" or "Please Fasten Seatbelts," which are not limited by this invention). The driver can then observe the surroundings and selectively click on the desired parking space in the panoramic image displayed on the interface. Specifically, they can click on any point within the desired parking space, which is referred to as an "interior point" in this invention. It should also be noted that with the prevalence of smart parking lots, some parking spaces have been equipped with smart ground locks with communication capabilities. Simultaneously, the vehicle is equipped with a pairing Bluetooth module. After the smart ground lock shakes and unlocks, the smart ground lock can serve as the "interior point" in the user's desired parking space. This means that the invention is not limited to the driver providing the interior point by clicking.

[0047] Next, the parking system uses the inner point as its geometric center and defines a rectangular area around it as a first region of interest. When defining the rectangular area, its length and width can be greater than or equal to the dimensions of two preset standard parking spaces. The purpose of designing this first region of interest is to limit the object of the parking space recognition algorithm to within the rectangular area, rather than searching for parking spaces globally.

[0048] Then, a second ROI is identified from the first ROI. The second ROI herein refers to the processing result output by the parking space recognition algorithm (a standard parking space containing the aforementioned inlier point), which can be released to the driver after post-processing (it should be noted that the inlier point is merely the geometric center of the first ROI and not necessarily the geometric center of the second ROI). Furthermore, the second ROI can have different types, such as vertical, horizontal, or inclined. The type of the second ROI itself can be identified by the parking space algorithm or customized by the driver. If the type of the second ROI (vertical, horizontal, or inclined) cannot be determined, the vertical parking space can be prioritized by default. The present invention is not limited to this. However, it should be emphasized that, since an arbitrary point within the desired parking space and the first ROI surrounding the inlier point are provided, the parking space recognition logic of the present invention differs from existing processing concepts and instead searches for a target within a rectangular area surrounding the inlier point, with the target requiring the inlier point to be included. Precisely because of this difference in parking space detection logic, a matching feature description method can be employed according to the present invention to train a dedicated parking space recognition model.

[0049] Specifically, the parking space identification strategy adopted by the present invention is to obtain the target parking space by lowering the confidence threshold used for verification of the parking space identification output result according to a preset gradient. It can be explained here that, usually in target identification schemes using machine learning or deep learning algorithms, a confidence threshold is preset in order to improve the target recognition accuracy. If the confidence of the detected target is lower than the threshold, it will be filtered out by the post-processing algorithm. It is based on this that if the second region of interest output is not successfully released, it indicates that the confidence verification has not passed. At this time, the current confidence threshold can be lowered once, and then the second region of interest that meets the conditions can be found. This design concept is, on the one hand, to ensure that the second region of interest can be released correctly and quickly, and on the other hand, to make it more likely that special parking areas that are not easily detected by existing algorithms will be released.

[0050] It should be pointed out here that the parking space recognition algorithm itself is not the focus of the present invention. In actual operation, conventional existing architectures such as machine learning-based parking space recognition models and parking obstacle avoidance models can be used to perform targeted training of the algorithm model used in the present invention. As mentioned above, for different types of parking spaces, there are also corresponding existing algorithms that can detect vertical parking spaces, horizontal parking spaces or inclined parking spaces, etc., which will not be elaborated in the present invention.

[0051] Continuing from the previous section, if there is still no releasable second ROI when the confidence threshold drops to a preset critical value, a pending parking area is provided to replace the second ROI and released to the user. Specifically, this process can utilize color gradient changes, edge detection, or line extraction algorithms within the first ROI to search for several edges (the number can be limited to two) adjacent to the interior point. Using the found edges, a pending parking area of ​​equal size and shape to the preset standard parking space is delineated and released to the driver. It can be understood that this concept utilizes edges and interior points to constrain a parking area, effectively replacing the output of the recognition algorithm. This allows for a reliable parking area to be provided for user selection even with fewer features (as long as there are no obstacles interfering).

[0052] After releasing the second area of ​​interest or the pending parking area to the driver, since the self-selected parking space function is currently activated, the driver can either directly lock the area as the target parking space; or, after being authorized, modify the position and posture of the pending parking area by sliding, rotating, etc., with the pending parking area as a reference.

[0053] When there are significant obstacles near the inner point, which causes the parking space recognition algorithm to be unable to output the second region of interest, the second region of interest is abandoned and the driver can be prompted that there is an obstacle and to select a parking space again.

[0054] Finally, after the user accepts the released second region of interest (which, of course, may also include the previously determined pending parking area), the predicted parking position and direction within the second region of interest can be determined based on the vehicle's head orientation and the aforementioned type of the second region of interest. Point-to-point parking motion planning can then be performed based on the predicted parking position and direction. It should be noted that the predicted parking position refers to a directional area that can essentially be locked by two quantities: the vehicle's rear axle center point and the vehicle's head orientation. The position of the rear axle center point can be constrained based on the scope of the second region of interest and the position of obstacles such as limit rods, thereby enabling point-to-point motion planning. Furthermore, the predicted parking position can be dynamically constrained based on the results of each segment of motion control during the actual parking process. That is, after each segment of the parking path is executed, the predicted final parking position is updated using the adjusted current vehicle posture. This is not the focus of the present invention and will not be discussed in detail.

[0055] In summary, the main design concept of the present invention is to formally eliminate the process of global search for parking spaces in the parking lot, and also weaken the requirements for real-time parking space identification. While reducing the operating pressure of the system, the saved computing power can be used to design more complex functions for assisted parking. Specifically, an inner point is obtained in the desired parking space selected by the user, a first region of interest is delineated based on the inner point, and a second region of interest containing the inner point is identified. The second region of interest that meets the confidence condition is released to the user, or the confidence threshold is lowered to detect the second region of interest again; after the user accepts and selects the released second region of interest, parking path planning is performed based on the second region of interest. The present invention changes the underlying idea of ​​traditional parking space identification, from global parking space identification to target parking space detection in the area surrounding a fixed point in the self-selected parking space, thereby reducing a large number of features involved in parking space identification, thereby being able to identify parking spaces that are more difficult to identify.

[0056] In the embodiment of the present invention, "at least one" refers to one or more, and "more" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent the existence of A alone, the existence of A and B at the same time, and the existence of B alone. A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b and c can be represented by: a, b, c, a and b, a and c, b and c, or a, b and c, where a, b, c can be single or multiple.

[0057] The above describes in detail the structure, features and effects of the present invention based on the embodiments shown in the drawings, but the above is only a preferred embodiment of the present invention. It should be noted that the technical features involved in the above embodiments and their preferred modes can be reasonably combined and matched into a variety of equivalent schemes by those skilled in the art without departing from or changing the design ideas and technical effects of the present invention; therefore, the scope of implementation of the present invention is not limited to what is shown in the drawings. Any changes made in accordance with the concept of the present invention, or modifications to equivalent embodiments with equivalent changes, which still do not exceed the spirit covered by the description and drawings, should be within the scope of protection of the present invention.

Claims

1. An automatic parking method based on self-selected parking space, characterized in that: include: After the self-selected parking space function of the parking system is activated, receiving an inner point of a desired parking space selected by the user in the panoramic image, wherein the inner point is any point selected by the user on the desired parking space; Taking the inner point as the geometric center, a rectangular range is set around it to obtain a first region of interest, which is used to limit the object of the parking space recognition algorithm to within the rectangular range; Invoking a preset parking space recognition algorithm to detect a second region of interest for parking within the first region of interest; wherein the second region of interest includes the interior point; If, after this round of detection, the second region of interest output by the parking space recognition algorithm meets the current confidence threshold, it is released to the user; the confidence threshold is used to verify the confidence of the output result of the parking space recognition algorithm; If the second ROI output by the parking space recognition algorithm is not released to the user after this round of detection, lowering the confidence threshold and detecting the second ROI again from within the first ROI; If the parking space recognition algorithm does not output the second region of interest after this round of detection, the user is prompted to reselect a desired parking space; After the user accepts and selects a released second region of interest, a predicted parking position and direction within the selected second region of interest are determined based on the vehicle's front orientation and the posture type of the selected second region of interest, and parking path planning is performed based on the predicted parking position and direction.

2. The automatic parking method based on self-selected parking space according to claim 1, characterized in that: When the confidence threshold is reduced to a preset critical value, if the second region of interest has not been released, a pending parking area replacing the second region of interest is allocated to the user.

3. The automatic parking method based on self-selected parking space according to claim 2, characterized in that: The method for obtaining the pending parking area includes: Searching for a plurality of edge lines adjacent to the interior point in the first region of interest by using a color gradient change and / or edge detection and / or line extraction algorithm; The found edge line is used to delineate an area having the same size and shape as the preset standard parking space.

4. The automatic parking method based on self-selected parking space according to claim 1, characterized in that: The size of the rectangular range is greater than or equal to the size of two preset standard parking spaces.

5. The automatic parking method based on self-selected parking space according to claim 1, characterized in that: The posture types include vertical, horizontal, and tilted.

6. The automatic parking method based on self-selected parking space according to claim 5, characterized in that: If the second region of interest output by the parking space recognition algorithm meets the current confidence threshold, releasing it to the user includes: preferentially releasing the second region of interest of a vertical type.

7. The automatic parking method based on self-selected parking space according to any one of claims 1 to 6, characterized in that: Lowering the confidence threshold includes lowering the confidence threshold according to a preset gradient.

Images