A parking space detection method based on parking space marking points

Abstract

The present invention discloses a parking space detection method based on parking space marking points, comprising the following steps: step 1: marking the parking space entrance head marking points and the marking frame according to the parking space entrance; step 2: establishing a parking space detection model; step 3: pairing the marking points, and screening the parking space entrance according to the constraint conditions; step 4: based on the marked corner points, inferring the other two invisible corner points of the parking space to form a complete parking space, and determining the parking space type based on the parking space angle information; step 5: based on the inferred complete parking space, designing an occupancy classification m_DCNN model based on improved AlexNet; step 6: judging whether the parking space is vacant based on step 5; step 7: integrating parking space detection and occupancy classification information. The present invention combines parking space classification with marking point positioning to infer the precise positions of various parking spaces; designs the m_DCNN model, optimizes the network structure and computing performance, and further improves the accuracy of parking space detection.

Description

Technical Field

[0001] The present invention relates to the technical field of parking space detection, and more specifically, to a parking space detection method based on parking space marking points. Background Art

[0002] Currently, most vehicles are equipped with Around View Monitor (AVM) when they leave the factory. Using AVM for marker-based parking space detection is a hot topic of current research. However, due to the influence of complex visual environments such as light occlusion, intensity changes, and visual accuracy, the detection accuracy and speed of parking space detection using AVM still need to be further improved.

[0003] The current mainstream technical visual parking space detection algorithms are mainly divided into parking space detection algorithms based on traditional images and parking space detection algorithms based on deep learning. The former mainly detects the separation lines and entrance lines of parking spaces by combining line clustering and multi-view learning, and finally uses geometric features to identify parking spaces. However, these classic methods are designed based on the original line features of parking spaces, and their robustness is relatively weak for real complex environments. Most of the detection methods based on deep learning are based on dual deep learning network design for parking space detection. The CNN feature extraction process used twice is highly repetitive. Therefore, there is still room for improvement in the detection efficiency of parking space detection algorithms based on deep learning. Summary of the invention

[0004] In order to overcome the shortcomings of the prior art, the present invention proposes a parking space detection method based on parking space marking points. In terms of parking space detection, parking space classification is combined with marking point positioning, and the precise positions of various parking spaces are inferred using geometric clues. In terms of occupancy classification, this paper designs a deep convolutional network m_DCNN model, optimizes the network structure and computing performance, and further improves the accuracy of parking space detection.

[0005] To achieve the above object, the present invention provides the following technical solution: a parking space detection method based on parking space marking points, the method comprising the following steps:

[0006] Step 1: Mark the parking space entrance head mark point and marking box according to the parking space entrance;

[0007] Step 2: Establish a parking space detection model;

[0008] Step 3: Pair the marked points and select the parking space entrances according to the constraints;

[0009] Step 4: Based on the marked corner points, infer the other two invisible corner points of the parking space to form a complete parking space, and determine the parking space type based on the parking space angle information;

[0010] Step 5: Based on the inferred complete parking space, design an occupancy classification m_DCNN model based on improved AlexNet;

[0011] Step 6: Determine whether the parking space is vacant based on step 5;

[0012] Step 7: Integrate parking space detection and occupancy classification information.

[0013] In step 1, the head mark point and the marking box are both marked at the entrance of the parking space. This method combines parking space positioning and mark point positioning;

[0014] The parking space head consists of parameters such as p(x, y), w1, h1, etc., which are obtained by calculating the coordinates of paired marking points using formulas (1)-(3). For each "T-type" and "L-type" marking point p i , its border is a width of w1, height of h1, p i A rectangle centered on

[0015]

[0016]

[0017]

[0018] Where p(x,y) is the center coordinate of the parking space head annotation box; p1(x,y) and p2(x,y) are the coordinates of the paired marked points of the entrance line; w1 is the width of the bounding box, h1 is the height of the bounding box; w2 is the width of the marked point annotation box, h2 is the height of the marked point annotation box, Δw and Δh are hyperparameters that control the width and height of the bounding box;

[0019] In step 2, the parking space detection model is based on the improved yolov4 parking space detection algorithm, the backbone network is replaced by resnet50, and deep separable convolution is added in the three feature extraction processes;

[0020] Step 3: Based on the direct relationship between the acquired marker points and the annotation box, determine whether the marker points are paired successfully;

[0021] Step 4: Determine the parking space type and angle based on the annotation box, and infer the two invisible points based on the angle information and the paired marked points to infer the complete parking space. The parking space inference calculation formula is:

[0022]

[0023]

[0024] In the formula, the angle α i and d iThey represent the parking angle and parking space depth respectively, p1 and p2 are the detected pairs of marked corner points, p3 and p4 are two invisible corner points, and the two invisible corner points p3 and p4 of the parking space.

[0025] In a preferred embodiment, the step five designs an occupancy classification m_DCNN model based on an improved AlexNet deep learning network.

[0026] In a preferred embodiment, step six determines whether the parking space is vacant based on the occupancy classification m_DCNN model designed in step five.

[0027] In a preferred embodiment, step seven integrates parking space detection and occupancy classification information to lay the foundation for subsequent path planning and tracking control.

[0028] Compared with the prior art, the technical effects and advantages of the present invention are as follows:

[0029] 1. A new parking space detection method based on panoramic images is proposed, which can quickly and accurately detect various types of parking spaces. By introducing the attention mechanism, the parking space proximity information is fully extracted to improve the parking space detection effect.

[0030] 2. A deep convolutional network m_DCNN model was designed to optimize the network structure and computing performance to further improve the accuracy of parking space detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is the operation flow chart of the present invention;

[0032] Figure 2 It is a structural block diagram of the present invention;

[0033] Figure 3 The bounding box of the parking space head and the marking point of the present invention;

[0034] Figure 4 It is the complete parking space inference of the present invention;

[0035] Figure 5 It is the m_DCNN network structure of the present invention;

[0036] Figure 6 A flow chart of detecting parallel and perpendicular parking space occupancy conditions disclosed in an example of the present invention. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0038] according to Figure 1-2 A parking space detection method based on parking space marking points is shown, the method comprising the following steps:

[0039] Step 1: Mark the parking space entrance head mark point and marking box according to the parking space entrance;

[0040] like Figure 3 As shown, Figure 3 The bounding box of the parking space head and the marking point of the present invention;

[0041] In step 1, the head mark point and the marking box are both marked at the entrance of the parking space. This method combines parking space positioning and mark point positioning;

[0042] The parking space head consists of parameters such as p(x, y), w1, h1, etc., which are obtained by calculating the coordinates of paired marking points using formulas (1)-(3). For each "T-type" and "L-type" marking point p i , its border is a width of w1, height of h1, p i A rectangle centered on

[0043]

[0044]

[0045]

[0046] Where p(x,y) is the center coordinate of the parking space head annotation box; p1(x,y) and p2(x,y) are the coordinates of the paired marked points of the entrance line; w1 is the width of the bounding box, h1 is the height of the bounding box; w2 is the width of the marked point annotation box, h2 is the height of the marked point annotation box, Δw and Δh are hyperparameters that control the width and height of the bounding box;

[0047] Step 2: Establish a parking space detection model; the parking space detection model is based on the improved yolov4 parking space detection algorithm, replacing the backbone network with resnet50, and adding deep separable convolution in the three feature extraction processes to improve the model detection speed;

[0048] Step 3: Pairing of marked points: Filtering the parking space entrances according to the constraints; Determining whether the marked points are successfully paired based on the direct relationship between the acquired marked points and the annotation boxes;

[0049] Step 4: Based on the marked corner points, infer the other two invisible corner points of the parking space to form a complete parking space, and determine the parking space type based on the parking space angle information;

[0050] like Figure 4 As shown, Figure 4 The complete parking space inference of the present invention; the parking space type and angle are determined based on the annotation box, and two invisible points are inferred based on the angle information and the paired marked points to infer the complete parking space. The parking space inference calculation formula is:

[0051]

[0052]

[0053] In the formula, the angle α i and d i Respectively represent the parking angle and parking space depth, p1, p2 are detected pairs of marked corner points, p3, p4 are two invisible corner points, two invisible corner points p3, p4 of the parking space;

[0054] Step 5: Based on the inferred complete parking space, design an occupancy classification m_DCNN model based on improved AlexNet;

[0055] like Figure 5 As shown, Figure 5 It is the m_DCNN network structure of the present invention;

[0056] Step 6: Determine whether the parking space is vacant based on the occupancy classification m_DCNN model designed in step 5;

[0057] Step 7: Integrate parking space detection and occupancy classification information to lay the foundation for subsequent path planning and tracking control.

[0058] In summary, the present invention provides a parking space detection method based on parking space marking points. When in use, firstly, in terms of parking space detection, parking space classification is combined with marking point positioning, and the precise positions of various parking spaces are inferred using geometric clues; in terms of occupancy classification, this paper designs a deep convolutional network m_DCNN model to optimize the network structure and computing performance, so as to further improve the accuracy of parking space detection; it can quickly and accurately detect various types of parking spaces, and fully extract parking space proximity information by introducing an attention mechanism to improve the parking space detection effect; a deep convolutional network m_DCNN model is designed to optimize the network structure and computing performance, so as to further improve the accuracy of parking space detection.

[0059] Finally, a few points should be explained: First, in the description of this application, it should be noted that, unless otherwise specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, which may refer to mechanical connection or electrical connection, or internal communication between two components, or direct connection. "upper", "lower", "left", "right", etc. are only used to indicate relative positional relationships. When the absolute position of the object being described changes, the relative positional relationship may change;

[0060] Secondly: In the drawings of the embodiments disclosed in the present invention, only the structures related to the embodiments disclosed in the present invention are involved, and other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of the present invention can be combined with each other;

[0061] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A parking space detection method based on parking space marking points, characterized in that: The method comprises the following steps: Step 1: Mark the parking space entrance head mark point and marking box according to the parking space entrance; Step 2: Establish a parking space detection model; Step 3: Pair the marked points and select the parking space entrances according to the constraints; Step 4: Based on the marked corner points, infer the other two invisible corner points of the parking space to form a complete parking space, and determine the parking space type based on the parking space angle information; Step 5: Based on the inferred complete parking space, design an occupancy classification m_DCNN model based on improved AlexNet; Step 6: Determine whether the parking space is vacant based on step 5; Step 7: Integrate parking space detection and occupancy classification information; In step 1, the head mark point and the marking box are both marked at the entrance of the parking space. This method combines parking space positioning and mark point positioning; The parking space head consists of parameters such as p(x,y), w1, h1, etc., which are obtained by calculating the coordinates of paired marking points using formulas (1)-(3). For each "T-type" and "L-type" marking point p i , its border is a width of w1, height of h1, p i A rectangle centered on Where p(x,y) is the center coordinate of the parking space head annotation box; p1(x,y) and p2(x,y) are the coordinates of the paired marked points of the entrance line; w1 is the width of the bounding box, h1 is the height of the bounding box; w2 is the width of the marked point annotation box, h2 is the height of the marked point annotation box, Δw and Δh are hyperparameters that control the width and height of the bounding box; In step 2, the parking space detection model is based on the improved yolov4 parking space detection algorithm, the backbone network is replaced by resnet50, and deep separable convolution is added in the three feature extraction processes; Step 3: Based on the direct relationship between the acquired marker points and the annotation box, determine whether the marker points are paired successfully; Step 4: Determine the parking space type and angle based on the annotation box, and infer the two invisible points based on the angle information and the paired marked points to infer the complete parking space. The parking space inference calculation formula is: In the formula, the angle α i and d i They represent the parking angle and parking space depth respectively, p1 and p2 are the detected pairs of marked corner points, p3 and p4 are two invisible corner points, and the two invisible corner points p3 and p4 of the parking space.

2. The parking space detection method based on parking space marking points according to claim 1, characterized in that: The step five designs an occupancy classification m_DCNN model based on an improved AlexNet deep learning network.

3. The parking space detection method based on parking space marking points according to claim 1, characterized in that: The step six determines whether the parking space is vacant based on the occupancy classification m_DCNN model designed in step five.

4. The parking space detection method based on parking space marking points according to claim 1, characterized in that: The step seven integrates parking space detection and occupancy classification information, laying the foundation for subsequent path planning and tracking control.

Images