KR20210087070A - Parking space detection method, device and electronic device - Google Patents

Abstract

A method, apparatus and electronic device for detecting a parking space, the method comprising: acquiring a parking space image (S101); inputting a parking space image to a neural network, and obtaining area information and/or corner point information of an empty parking space from the parking space image (S102); and determining a detection result of an empty parking space in the parking space image based on area information and/or corner point information of the empty parking space in the parking space image (S103).

Description

Parking space detection method, device and electronic device

Cross-referencing of related applications

The present invention is filed based on a Chinese patent application with an application number of 201910458754.4 and an filing date of May 29, 2019, and claims the priority of the Chinese patent application, all contents of the Chinese patent application are incorporated herein by reference.

The present invention relates to artificial intelligence technology, and more particularly to a method, apparatus and electronic device for detecting a parking space.

With the improvement of people's living standards, automobiles have become an essential means of transportation in life, and with the development of computer vision, autonomous driving has come into the limelight. One key task in autonomous driving is the detection of a parking space, the purpose of which is to automatically find an empty space and park the vehicle in the empty space.

An embodiment of the present invention provides a method, apparatus and electronic device for detecting a parking space.

According to a first aspect, an embodiment of the present invention provides a method for detecting a parking space, the method comprising: acquiring a parking space image; inputting the parking space image to an instance segmentation neural network, and obtaining area information and/or corner point information of an empty parking space from the parking space image; and determining a detection result of an empty parking space in the parking space image based on area information and/or corner point information of the empty parking space in the parking space image.

According to a second aspect, an embodiment of the present invention provides an apparatus for detecting a parking space, the apparatus comprising:

a first acquiring module for acquiring a parking space image;

a processing module for inputting the parking space image into a neural network to obtain area information and/or corner point information of an empty parking space from the parking space image; and

and a determining module for determining a detection result of an empty parking space in the parking space image based on area information and/or corner point information of the empty parking space in the parking space image.

According to a third aspect, an embodiment of the present invention provides an electronic device, comprising: a memory in which a computer program is stored; and a processor executing the computer program to implement the method of detecting a parking space according to the first aspect.

According to a fourth aspect, an embodiment of the present invention provides a computer storage medium in which a computer program is stored, and when the computer program is executed, the method of detecting a parking space according to the first aspect is implemented.

The method, apparatus and electronic device for detecting a parking space provided in an embodiment of the present invention include inputting a parking space image to a neural network through acquiring a parking space image, and area information of an empty parking space in the parking space image and / or obtain corner point information; Based on area information and/or corner point information of an empty parking space in the parking space image, a detection result of an empty parking space in the parking space image is determined. That is, in the detection method of the embodiment of the present invention, area information and/or corner point information of an empty parking space can be obtained only by inputting the obtained parking space image to the neural network, and there is no need for prior image processing and the overall detection process This simple and short time required, based on the area information and/or corner point information of the empty parking space in the parking space image, determines the detection result of the empty parking space in the parking space image, thereby effectively improving the detection accuracy of the empty parking space improve

In order to more clearly describe the embodiments of the present invention or the technical solutions of the prior art, a brief introduction is made to the accompanying drawings used in the description of the embodiments or the prior art below, and the drawings described below are It is obvious that only some embodiments described in, those skilled in the art can further obtain other drawings according to these drawings, under the premise that the person skilled in the art does not strive to create inventive step.
1 is a flow schematic diagram 1 of a parking space detection method provided in an embodiment of the present invention.
2 is an exemplary view of a parking space.
Figure 3 is a flow schematic diagram 2 of the parking space detection method provided in the embodiment of the present invention.
4A is an exemplary diagram of a parking space training image used in an embodiment of the present invention.
4B is an image after key point tagging is performed with respect to FIG. 4A.
5 is a training flow diagram of an instance partitioning network referred to in an embodiment of the present invention.
6 is a structural schematic diagram of an instance partitioning network mentioned in an embodiment of the present invention.
7 is a schematic diagram of a parking space detection result mentioned in an embodiment of the present invention.
8 is a structural schematic diagram 1 of a detection device for a parking space provided in an embodiment of the present invention.
9 is a structural schematic diagram 2 of a detection device for a parking space provided in an embodiment of the present invention.
10 is a structural schematic diagram 3 of a detection device for a parking space provided in an embodiment of the present invention.
11 is a structural schematic diagram of an electronic device provided in an embodiment of the present invention.

In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be accurately and completely described below with reference to the drawings in the embodiments of the present invention. It is clear that the described embodiments are merely some embodiments of the present invention and not all embodiments. Based on the embodiments in the present invention, all other embodiments obtained by those skilled in the art without exerting effort to create inventive step shall all fall within the protection scope of the embodiments of the present invention.

Embodiments of the present invention may be applied to electronic devices such as terminal devices, computer systems, servers, etc., which may work with numerous other general-purpose or dedicated computing system environments or configurations. Well-known terminal devices, computing systems, environments, and/or personal computer systems, server computer systems, thin clients, rich clients, handheld and laptop devices, microprocessors and Systems based on Central Processing Units (CPUs) and Graphics Processing Units (GPUs), in-vehicle systems, set-top boxes, programmable consumer electronics, network personal computers, small computer systems, large computer systems and any of the above It applies to configured examples including, but not limited to, at least one of a distributed cloud computing technology environment comprising a system.

Electronic devices, such as terminal devices, computer systems, and servers, may be described in a general language environment of computer system executable commands (eg, program modules) executed by the computer system. Typically, program modules may include example programs, programs, target programs, components, logic, data structures, etc., which perform specified tasks or specified abstract data types. The computer system/server may be implemented in a distributed cloud computing environment, where tasks are executed via remote processing devices of a communication network link. In a distributed cloud computing environment, program modules may be located in a local or remote computing system storage medium including a storage device.

In an exemplary application, the electronic device may be installed in a vehicle and connected to a reversing system, which assists the reversing system to park the vehicle in an empty parking space. In another exemplary application, the electronic device is connected with the auxiliary driving system, and the electronic device sends the obtained detection result of the empty parking space to the auxiliary driving system, so that the auxiliary driving system determines the vehicle according to the detection result of the empty parking space It controls driving, for example, so that the vehicle is parked in an empty parking space. Optionally, the electronic device may be part or all of an auxiliary driving system, or part or all of a reverse system. Optionally, the electronic device may be connected to a control system of another vehicle according to actual demand, but the embodiment of the present invention is not limited thereto.

Hereinafter, the technical solutions of the present invention will be described in detail with reference to specific embodiments. Hereinafter, several specific embodiments may be combined with each other, and some embodiments may not be repeatedly described for the same or similar concept or process.

1 is a flow schematic diagram 1 of a method for detecting a parking space provided in an embodiment of the present invention. 1 , the method of this embodiment may include the following steps.

Step S101: Acquire a parking space image.

The present embodiment exemplifies that the execution subject is an electronic device, and the electronic device may include, but not be limited to, a smartphone, a computer, a vehicle-mounted system, and the like.

Optionally, FIG. 2 is an exemplary view of a parking space. The electronic device of this embodiment further includes a camera, and the camera can capture the driving environment of the vehicle, for example, a road on which the vehicle travels with the camera. A parking space image can be obtained by taking an image of a surrounding parking space, and the parking space image is sent to the processor of the electronic device, so that the processor performs the method of this embodiment, and the detection result of an empty parking space in the parking space image to get

Optionally, the electronic device of the present embodiment may be connected to an external camera, and may acquire a parking space image by photographing the driving environment of the vehicle with the external camera.

Optionally, the imaging module of the camera of the embodiment of the present invention may be a complementary metal oxide semiconductor (CMOS) or a charge coupled device (CCD), but is not limited thereto.

Step S102: Input the parking space image to a neural network to obtain area information and/or corner point information of an empty parking space from the parking space image.

The neural network of the embodiment of the present invention includes, but is not limited to, a back propagation (BP) neural network, a radial basis function (RBF) neural network, a sensor neural network, a linear neural network, a feedback neural network, and the like.

Optionally, the neural network may implement instance segmentation, where instance segmentation refers to not only proceeding with pixel-level classification, but also discriminating different implementations on the basis of specific categories. For example, in the parking space image, there are a plurality of empty parking spaces A, B, and Bottle, and instance segmentation may identify these three empty parking spaces as different objects.

An embodiment of the present invention may detect area information of an empty parking space and/or corner point information of an empty parking space from a parking space image through a neural network. For example, by training a neural network through a parking space image set with area information of an empty parking space and/or corner point information of an empty parking space in advance, the neural network extracts area information and/or corner point information of an empty parking space to learn the ability to do, thereby inputting the parking space image shown in Fig. 2 into the neural network, and area information and/or corner point information of an empty parking space in the parking space image obtained by the neural network through processing on the parking space image can be printed out.

Here, the area information of the empty parking space may include information such as the location and size of the empty parking space; The corner point information includes position information of a corner point of an empty parking space.

Optionally, the corner point information of the empty parking space may include corner point information of at least three corner points of the empty parking space. Since the parking space is usually rectangular, area information of the empty parking space may be determined according to the corner point information of at least three corner points of the empty parking space.

The detection method of the embodiment of the present invention can obtain area information and/or corner point information of an empty parking space only by inputting the obtained parking space image to the neural network, and does not require prior image processing, and the overall detection process This is simple, and the time required is short.

Step S103, a detection result of an empty parking space in the parking space image is determined based on the area information and/or corner point information of the empty parking space in the parking space image.

In one embodiment, the area information of the empty parking in the parking image may be the detection result of the empty parking space. In another embodiment, corner point information of an empty parking in the parking image may be used as a detection result of an empty parking space.

In another embodiment, the step of determining a detection result of an empty parking space in the parking space image based on the area information and corner point information of the empty parking space in the parking space image comprises: an empty parking space in the parking space image and determining a detection result of an empty parking space in the parking space image by merging the area information and the corner point information. Here, the step of merging the area information and corner point information of the empty parking space and determining the detection result of the empty parking space includes, but is not limited to, the following method.

Method 1: Determine parking space area information constituted by corner point information of an empty parking space in the parking space image; The parking space area information constituted by the area information of the empty parking space and the corner point information in the parking space image is merged, and the detection result of the empty parking space in the parking space image is determined based on the area information after merging. For example, the parking space area information surrounded by the corner point information of the empty parking space is displayed as the parking space area information 1, the area information of the empty parking space in the parking space image is displayed as the area information 2, and the parking space area information 1 and area information 2 are merged to obtain one area information 3, for example, the average value of parking space area information 1 and area information 2 is taken as area information 3, and area information 3 after merging is used as an empty parking space image in the parking space image. Let it be the result of detection of space.

Method 2: Determine corner point information of area information of an empty parking space in the parking space image; The corner point information of the empty parking space and the corner point information of the area information are merged in the parking space image, and the detection result of the empty parking space in the parking space image is determined based on the corner point information after merging. For example, corner points corresponding to area information of an empty parking space in the parking space image are displayed as corner points a1, corner points a2, corner points a3, and corner points a4, respectively, and corner points of empty parking spaces in the parking space image A corner point corresponding to the information is denoted as a corner point b1, a corner point b2, a corner point b3, and a corner point b4, respectively, wherein the corner point a1, the corner point a2, the corner point a3, the corner point a4 are respectively the corner point b1, The corner point b2, the corner point b3, and the corner point b4 correspond one-to-one, so that the corresponding two corner points can be merged into one corner point, for example, the corner point a1 and the corner point b1 into one By merging into the corner point ab1, new corner point information can be obtained, and this new corner point information is used as a detection result of an empty parking space in the parking space image.

An embodiment of the present invention can improve the detection accuracy of an empty parking space by merging the area information and corner point information of the empty parking space in the parking space image to determine the detection result of the empty parking space in the parking space image .

Optionally, the method according to an embodiment of the present invention may perform the above steps when the vehicle finds a parking space. For example, when the autonomous driving system arrives at a destination or receives a command to find a parking space sent from outside, the autonomous driving system controls the electronic device to operate. In the above scene, if the electronic device has a camera, the processor in the electronic device controls the camera to take an image of the parking space around the vehicle, and if the electronic device does not have a camera, the electronic device sends a shooting command to an external camera By sending it, the image of the parking space around the vehicle photographed by the camera is sent to the electronic device. After the electronic device acquires the parking space image, processing is performed on the parking space image to detect a detection result of an empty parking space in the parking space image. For example, the electronic device inputs the obtained parking space image to the neural network, outputs area information and/or corner point information of an empty parking space in the parking space image through processing of the neural network, and then outputs the empty parking space image in the parking space image. Accurate detection of an empty parking space is implemented by determining a detection result of an empty parking space in a parking space image based on area information and/or corner point information of the parking space.

Optionally, the electronic device may also be connected to the autonomous driving system, and send the detection result of the empty parking space to the autonomous driving system, wherein the autonomous driving system controls the vehicle according to the detection result of the empty parking space to Please park in an empty parking space.

The parking space detection method provided in the embodiment of the present invention includes inputting a parking space image to a neural network through acquiring a parking space image, and obtaining area information and/or corner point information of an empty parking space from the parking space image. obtain; Based on area information and/or corner point information of an empty parking space in the parking space image, a detection result of an empty parking space in the parking space image is determined. That is, in the detection method of the embodiment of the present invention, area information and/or corner point information of an empty parking space can be obtained only by inputting the obtained parking space image to the neural network, and there is no need for prior image processing and the overall detection process This simple and short time required, based on the area information and/or corner point information of the empty parking space in the parking space image, determines the detection result of the empty parking space in the parking space image, thereby improving the detection accuracy of the empty parking space make it

In some embodiments, in order to detect an incomplete empty parking space in a parking space image, the method of an embodiment of the present invention includes inputting the parking space image to the neural network in step S102, so that the area of the empty parking space in the parking space image Before the step of obtaining information and/or corner point information, the method further includes the following steps.

Step S102a: Extend from the peripheral edge of the parking space image outward by a preset value.

Optionally, the preset value is less than or equal to half the length of the parking space.

Exemplarily, as shown in FIGS. 4A and 4B , if it is hypothesized that FIG. 4A is a parking space image obtained by an electronic device, the parking space image includes two empty parking spaces: an empty parking space 1 and an empty parking space 2 included, wherein the parking space image only includes a part of the empty parking space 2 . To detect the empty parking space 2, the peripheral edge of the parking space image shown in Fig. 4A is extended outward by a preset value, and as shown in the black border in Fig. 4B, the result shown in Fig. 4B is obtained. Thereby, the visual range of the parking space image can be enlarged, and some empty parking spaces located outside the parking space image can be detected, thereby further increasing the accuracy of parking space detection.

At this time, inputting the parking space image to the neural network in step S103 to obtain area information and/or corner point information of an empty parking space in the parking space image may be replaced by step S103a.

Step S103a: Input the expanded parking space image to the neural network to obtain area information and/or corner point information of an empty parking space from the parking space image.

For example, FIG. 4B may be input to the neural network to detect area information of empty parking space 1 and empty parking space 2 and/or corner point information of empty parking space 1 and empty parking space 2 in FIG. 4B .

In the method of the embodiment of the present invention, before detecting the area information and/or corner point information of an empty parking space in the parking space image, the parking space after extending from the peripheral edge of the parking space image outward by a preset value The image is fed back into the neural network, which can detect some empty parking spaces located outside the parking space image, further improving the accuracy and practicality of parking space detection.

3 is a flow schematic diagram 2 of a parking space detection method provided in an embodiment of the present invention, and on the basis of the embodiment, the method of the embodiment of the present invention may further include a process of training a neural network, in FIG. As shown, the training process includes the following steps.

Step S201: Acquire a plurality of parking space training images.

The plurality of parking space training images may be obtained by an electronic device from a database, may be taken by the electronic device in the past, and in an embodiment of the present invention, for the step of the electronic device acquiring a plurality of parking space training images The specific process is not limited.

Here, each parking space image includes one or a plurality of empty parking spaces, for example, FIG. 4B is a parking space training image, where it includes an empty parking space 1 and an empty parking space 2 .

Optionally, the parking space training image may be an image collected using a wide-angle camera, wherein the image has a prescribed torque. When the neural network is trained with the parking space training image taken using the wide-angle camera, the neural network after training can predict the parking space image taken for different views, and furthermore, on the premise of ensuring the prediction accuracy, the parking space image reduced shooting requirements.

The key point of an empty parking space may include a point of a parking space sideline, a corner point of an empty parking space, or an intersection of two diagonals of an empty parking space, and based on these key points, the area and location of the empty parking space are determined can be obtained accurately.

The parking space training image includes tagging information for key point information of an empty parking space. For example, as shown in FIG. 4B , key points of empty parking space 4 are tagged as key point 1 , key point 2 , key point 3 and key point 4 . It should be explained that Fig. 4B is one tagging method of the key point of the empty parking space 1, the key point of the empty parking space 1 includes, but is not limited to, the four key points, and the key point of the empty parking space 1 The specific number and selection method of are determined according to actual demand, and the embodiment of the present invention is not limited thereto.

Step S202: Train the neural network by using the plurality of parking space training images.

A plurality of parking space training images containing tagging information for key point information of an empty parking space are input to the neural network, and based on the difference between the detection result input to the neural network and the tagging information for the key point information of an empty parking space to adjust the network parameters of the neural network to complete the training of the neural network.

In one possible embodiment, the neural network extends the peripheral edge of the used parking space training image out by a preset value in the training process of the neural network to implement detection of an incomplete empty parking space in the parking space image. . Optionally, the preset value is less than or equal to half the length of the parking space.

Continuing to refer to FIG. 4B , the peripheral edge of the parking space training image shown in FIG. 4B is extended outward by a preset value, the preset value being less than or equal to half the length of the parking space, and the black edge in FIG. 4B is suggested as Thereby, the visual range of the parking space training image can be enlarged, and the purpose is to detect some empty parking spaces located outside the captured parking space image when the neural network after training detects a subsequent parking space.

After extending the peripheral edge of the parking space training image shown in Fig. 4b outward by a preset value, the empty parking space 2 in Fig. 4b can be displayed completely, thereby tagging the key points of the empty parking space 2, and , for example, the key points where empty parking space 2 is tagged are key point 11, key point 12, key point 13, key point 14, key point 15 and key point 16, where key point 14 is located within the extended area. do. It should be explained, that Fig. 4b is only one possible tagging method of key point information of empty parking space 2, the key points of empty parking space 2 include, but are not limited to, the above 6 key points, empty parking space 2 The specific number and selection method of the key points of are determined according to actual demand, and the embodiment of the present invention is not limited thereto.

Optionally, the number and selection method of the key points of the empty parking space 1 and the empty parking space 2 may be the same or different, and each key point of the empty parking space 1 is sequentially connected so that the enclosed area is an empty parking space It is the area of space 1, and each key point of empty parking space 2 is sequentially connected to ensure that the enclosed area is the area of empty parking space 2.

In FIG. 4B , key points 3 and 4 of empty parking space 1 overlap with key points 13 and 11 of empty parking space 2 .

Optionally, the key point information of the empty parking space in the parking space training image includes at least one corner point information of the empty parking space, wherein the corner point corresponding to each corner point information is two pieces of the empty parking space. It is the intersection of the sidelines.

In an embodiment of the present invention, by extending the peripheral edge of the parking space training image outward by a preset value and supplementing the incomplete empty parking space, the standard of key point information of the incomplete empty parking space in the parking space training image information, and training the neural network using these parking space training images, the neural network after training can predict that there is no picture of a completely empty parking space in the parking space image, and the overall performance of parking space detection accuracy can be improved.

On the basis of the training embodiment of the neural network, in one possible implementation manner, as shown in Fig. 5 , using the plurality of parking space training images in the step S202 to train the neural network is specifically the following may include steps.

Step S301: Obtain area information composed of corner point information of an empty parking space in the parking space training image and key point information of an empty parking space in the parking space training image.

Illustratively, taking the empty parking space 2 in FIG. 4B as an example, six key points of the empty parking space 2 include four corner points, for example, the tagging of the four corner points is indicated by 1, and another Mark the key point as 0, so that the 6 key point information of empty parking space 2 is {"kpts":[[1346.2850971922246, 517.6241900647948, 1.0], [1225.010799136069, 591.1447084233262,1.0], [1280.6479481641468, 666.6522678185745, 0.0] , [1300.5183585313175, 728.2505399568034, 1.0], [1339.2656587473002, 707.3866090712743, 0.0], [1431.6630669546437, 630.8855291576674, 1.0]]}.

obtaining area information constituted by the six key point information of the empty parking space 2, for example, arranging the six key points of the empty parking space 2 in a clockwise or counterclockwise direction, and then obtaining the enclosed area; , let the enclosed area be area information of the vacant parking space 2 . The area information of the empty parking space 2 may be used as the truth value of the area information in the detection process of the parking space.

Referring to the above method, area information composed of key point information of an empty parking space is obtained from each one of a plurality of parking space training images among a plurality of parking space training images to form a ground truth of an empty parking space .

In addition, continuing taking the empty parking space 2 in Fig. 4B as an example, four corner point information of the empty parking space 2 is obtained from the six key point information of the empty parking space 2, {"corner point":[[1346.2850971922246 , 517.6241900647948, 1.0], [1225.010799136069, 591.1447084233262,1.0], [1300.5183585313175, 728.2505399568034, 1.0], [1431.6630669546437, 630.8855291576674, 1.0]}. . The four corner point information constitutes the corner point truth value of the empty parking space 2 .

Referring to the above method, the corner point truth value of the empty parking space is formed from the corner point information of the empty parking space in each one of the parking space training images among the plurality of parking space training images.

Step S302: Train the neural network by using corner point information and area information of an empty parking space in the parking space training image and the parking space training image.

Referring to FIG. 6 , FIG. 6 is a structural schematic diagram of a neural network mentioned in the embodiment of the present invention, and the neural network mentioned in the embodiment of the present invention includes, but is not limited to, the neural network shown in FIG. 6 .

6 , the neural network may include an instance partitioning layer, and the instance partitioning layer acquires area information of an empty parking space.

Optionally, the neural network of the embodiment of the present invention is a neural network of a mask-RCNN structure, and as shown in FIG. 6 , the neural network includes a Feature Pyramid Networks (FPN) detection base unit and a region convolutional neural network (Region). CNN, RCNN) further comprising a positioning layer, wherein an output end of the FPN detection base unit is connected to an input end of the RCNN position return layer, and an output end of the RCNN position return layer is connected to an input end of the instance partitioning layer. Accordingly, the FPN detection base unit can detect the detection frame of the empty parking space in the parking space training image, as is the rectangular frame shown in FIG. 6 . Then, the detection frame of the detected empty parking space is input to the RCNN position return layer, and the RCNN position return layer fine-tunes the detection frame of the empty parking space detected by the FPN detection base unit. The RCNN location return layer inputs the unadjusted detection frame of the empty parking space to the instance partitioning layer, and the instance partitioning layer divides the area information of the empty parking space, for example, as in the white area of FIG. 7 .

Optionally, the instance partitioning layer is formed by stacking a series of convolutional layers or pooling layers according to a preset order.

Continuing to refer to FIG. 6 , the neural network may further include a key point detection layer, wherein the key point detection layer acquires corner point information of an empty parking space.

Optionally, the input end of the key point detection layer is connected to the output end of the RCNN position return layer, and the RCNN position return layer inputs the unadjusted detection frame of the empty parking space to the key point detection layer, and detects the key point. The floor outputs corner point information of an empty parking space, for example, the same as the black corner point in FIG. 7 . It should be explained that in FIG. 7 , one sideline of two adjacent empty parking spaces overlaps, so that two corner points of the other two empty parking spaces overlap.

According to the method, through the area information and corner point information of the empty parking space predictable by the neural network, the area information of the empty parking space predicted again and the key point information of the empty parking space in the parking space training image obtained in the above step Comparing the constituting area information, and comparing the predicted corner point information of the empty parking space with the corner point information of the empty parking space in the parking space training image obtained in the above step, the parameters of the neural network are adjusted. The above steps are repeated until the number of training times of the neural network reaches a preset number or the prediction error of the neural network reaches a preset error.

The method of an embodiment of the present invention includes: obtaining area information constituted by corner point information of an empty parking space in the parking space training image and key point information of an empty parking space in the parking space training image; By using the corner point information and area information of the empty parking space in the parking space training image and the parking space training image, the neural network is trained, and the neural network after training receives area information and/or corner point information of the empty parking space to be able to predict accurately.

The detection method of any parking space provided in the embodiment of the present invention may be performed by any suitable device having data processing capability, including but not limited to a terminal device or a server. The detection method of any parking space provided in the embodiment of the present invention is performed by a processor, for example, the processor calls a corresponding instruction stored in the memory to detect any one parking space mentioned in the embodiment of the present invention. do the method No further description is given below.

A person skilled in the art will know that all or some steps of the method embodiments are completed by hardware related to program instructions, the above-described program may be stored in a computer-readable storage medium, and when the program is executed, the steps of including the method embodiments perform; It can be understood that the above-described storage medium includes a storage medium of various program codes, such as ROM, RAM, disk, or CD.

8 is a structural schematic diagram 1 of a detection device for a parking space provided in an embodiment of the present invention. As shown in Figure 8, the detection device 100 of the parking space of this embodiment,

a first acquiring module 110 for acquiring a parking space image;

a processing module 120 for inputting the parking space image into a neural network to obtain area information and/or corner point information of an empty parking space from the parking space image; and

and a determination module 130 that determines a detection result of an empty parking space in the parking space image based on area information and/or corner point information of the empty parking space in the parking space image.

The apparatus for detecting a parking space in an embodiment of the present invention can implement the technical solution of the above method embodiment, and its implementation principle and technical effect are similar, so that it is not further described herein.

In one possible embodiment, the determining module 130 determines the detection result of an empty parking space in the parking space image by merging the area information and corner point information of the empty parking space in the parking space image.

In another possible embodiment, the determining module 130 determines the parking space area information constituted by the corner point information of the empty parking space in the parking space image; In the parking space image, area information of an empty parking space and parking space area information constituted by the corner point information are merged, and a detection result of an empty parking space in the parking space image is determined.

9 is a structural schematic diagram of an apparatus for detecting a parking space provided in an embodiment of the present invention, and the apparatus 100 for detecting a parking space further includes an expansion module 140,

the expansion module 140 extends outward from the peripheral edge of the parking space image by a preset value, wherein the preset value is less than or equal to half the length of the parking space;

The processing module 120 inputs the expanded parking space image to the neural network to obtain area information and/or corner point information of an empty parking space from the parking space image.

The apparatus for detecting a parking space in an embodiment of the present invention can implement the technical solution of the above method embodiment, and its implementation principle and technical effect are similar, so that it is not further described herein.

10 is a structural schematic diagram 3 of a detection device for a parking space provided in an embodiment of the present invention, the detection device 100 for a parking space,

a second acquisition module 150 for acquiring a plurality of parking space training images; and

and a training module 160 for training the neural network by using the plurality of parking space training images, wherein the parking space training image includes tagging information for key point information of an empty parking space.

In one possible embodiment, the peripheral edge of the parking space training image extends outward by a preset value, wherein the preset value is less than or equal to half the length of the parking space.

In another possible embodiment, the key point information of an empty parking space in the parking space training image includes at least one corner point information of the empty parking space.

In another possible embodiment, the training module 160 obtains area information constituted by corner point information of an empty parking space in the parking space training image and key point information of an empty parking space in the parking space training image and; The neural network is trained by using corner point information and area information of an empty parking space in the parking space training image and the parking space training image.

Optionally, the parking space training image is an image captured by a wide-angle camera.

The apparatus for detecting a parking space in an embodiment of the present invention may implement the technical solution of the above method embodiment, and its implementation principle and technical effect are similar, so that it will not be further described herein.

11 is a structural schematic diagram of an electronic device provided in an embodiment of the present invention. As shown in FIG. 11 , the electronic device 30 of this embodiment is,

a memory 310 in which a computer program is stored; and

Since it includes a processor 320 that executes the computer program to implement the method of detecting the parking space, its implementation principle and technical effect are similar, so that it will not be further described herein.

In addition, when at least a part function of the method for detecting a parking space in the embodiment of the present invention is implemented as software, the embodiment of the present invention further provides a computer storage medium, which is a volatile or non-volatile computer storage medium, and the computer storage The medium stores the detecting computer software instructions of the parking space, which, when run in a computer, enable the computer to perform the various possible parking space detecting methods of the method embodiments. When the computer loads and executes the computer-executed instructions, all or part of the flow or function according to the embodiment of the present invention is followed. The computer instructions may be stored on a computer storage medium or transmitted from one computer storage medium to another, and the transmission may be performed wirelessly (eg, cellular communication, infrared, short-range radio, microwave, etc.). ) to another website, computer, server or data-centric. The computer storage medium may be any available computer-accessible medium or a data storage device including a server, a data center, etc. integrated into one or a plurality of available media. The usable medium may be a magnetic medium (eg, a floppy diskette, hard disk, magnetic tape), an optical medium (eg, a Digital Versatile Disc (DVD)), or a semiconductor medium (eg, a solid state disk). (Solid State Disk, SSD)) and the like.

In the above embodiment, all or part may be implemented in software, hardware, firmware, or any combination thereof. When implemented in software, it may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed in a computer, they generate, in whole or in part, the flow or function according to an embodiment of the present invention. The computer may be a general purpose computer, a dedicated computer, a computer network, or other programmable device. The computer instructions may be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another computer readable storage medium, for example, the computer instructions may be stored in one website; From a computer, server, or data center to another website, either wired (eg coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (eg infrared, wireless, microwave, etc.) , computer, server, or data-centric. The computer-readable storage medium may be any available medium that a computer can access or a data storage device including a server, a data center, etc. integrated in one or a plurality of available media. The usable medium may be a magnetic medium (eg, a floppy diskette, a hard disk, magnetic tape), an optical medium (eg, a Digital Versatile Disc (DVD)), or a semiconductor medium (eg, a solid state disk). (Solid State Disk, SSD)) and the like.

In an embodiment of the present invention, “at least one” refers to one or a plurality, and “plurality” refers to two or more than two. "And/or" describes the relationship of the object of association and means that three relationships exist. For example, A and/or B means three cases in which only A exists, A and B exist simultaneously, and only B exists. Here, A and B may be singular or plural. The sign “/” generally indicates that the related object is a relation of one “or”; The sign “/” in the formula indicates that the relation before and after is a relation of one “division”. “At least one term below” or a similar expression refers to any combination of these terms, including any combination of the single term or plural terms. For example, at least one term (one) of a, b, or c may represent a, b, c, ab, ac, bc, or abc, where a, b, c may be one or plural. may be

It can be understood that various numerical numbers referred to in the embodiments of the present invention are merely for convenience of description, and do not limit the scope of the embodiments of the present invention.

In various embodiments of the present invention, the size of the sequence number of each process does not mean before and after the execution order, and the execution order of each process should be determined according to its function and inherent logic, and the implementation process according to the embodiment of the present invention It should be understood that there is no limitation on the

Finally, each of the above embodiments is only for explaining the technical solutions of the present invention, and is not limited thereto. Although the present invention is described in detail with reference to each of the above-described embodiments, it should be understood by those skilled in the art that it is still a correction for the technical solutions described in each of the above-described embodiments, or equivalent replacement of partial or all technical features therein. . However, such correction or replacement does not deviate from the scope of the technical solutions of each embodiment of the present invention in essence of the corresponding technical solutions.

Claims (20)

A method for detecting a parking space, comprising:
acquiring a parking space image;
inputting the parking space image to a neural network to obtain at least one of area information and corner point information of an empty parking space from the parking space image; and
and determining a detection result of an empty parking space in the parking space image based on at least one of area information and corner point information of the empty parking space in the parking space image. According to claim 1,
The step of determining the detection result of the empty parking space in the parking space image based on the area information and the corner point information of the empty parking space in the parking space image,
and determining a detection result of an empty parking space in the parking space image by merging area information and corner point information of the empty parking space in the parking space image. 3. The method of claim 2,
The step of determining the detection result of the empty parking space by merging the area information and corner point information of the empty parking space in the parking space image,
determining parking space area information constituted by corner point information of an empty parking space in the parking space image; and
Merging the area information of the empty parking space and the parking space area information constituted by the corner point information in the parking space image, and determining the detection result of the empty parking space in the parking space image based on the area information after merging A method of detecting a parking space comprising a. 4. The method according to any one of claims 1 to 3,
Before inputting the parking space image to a neural network to obtain at least one of area information and corner point information of an empty parking space in the parking space image, the method comprises:
extending out from the peripheral edge of the parking space image by a preset value, wherein the preset value is less than or equal to half the parking space length;
The step of inputting the parking space image to a neural network to obtain at least one of area information and corner point information of an empty parking space from the parking space image comprises:
and inputting the expanded parking space image to the neural network to obtain at least one of area information and corner point information of an empty parking space from the parking space image. 5. The method according to any one of claims 1 to 4,
The method is
acquiring a plurality of parking space training images; and
Using the plurality of parking space training images, further comprising the step of training the neural network,
The parking space training image is a parking space detection method including tagging information for key point information of an empty parking space. 6. The method of claim 5,
A method of detecting a parking space by extending the peripheral edge of the parking space training image outward by a preset value, wherein the preset value is less than or equal to half the length of the parking space. 7. The method of claim 5 or 6,
In the parking space training image, the key point information of the empty parking space includes at least one corner point information of the empty parking space. 8. The method of claim 7,
Using the plurality of parking space training images, training the neural network comprises:
obtaining area information composed of corner point information of an empty parking space in the parking space training image and key point information of an empty parking space in the parking space training image; and
and training the neural network by using corner point information and area information of an empty parking space in the parking space training image and the parking space training image. 9. The method according to any one of claims 5 to 8,
The parking space training image is an image taken with a wide-angle camera. A parking space detection device, comprising:
a first acquiring module for acquiring a parking space image;
a processing module for inputting the parking space image into a neural network to obtain at least one of area information and corner point information of an empty parking space from the parking space image; and
and a determining module configured to determine a detection result of an empty parking space in the parking space image based on at least one of area information and corner point information of the empty parking space in the parking space image. 11. The method of claim 10,
The determining module is a parking space detection device for merging area information and corner point information of an empty parking space in the parking space image to determine a detection result of an empty parking space in the parking space image. 12. The method of claim 11,
the determining module determines parking space area information constituted by corner point information of an empty parking space in the parking space image; Parking for merging the area information of the empty parking space and the parking space area information constituted by the corner point information in the parking space image, and determining the detection result of the empty parking space in the parking space image based on the area information after merging space detection device. 13. The method according to any one of claims 10 to 12,
The device is
an expansion module extending out from the peripheral edge of the parking space image by a preset value, wherein the preset value is less than or equal to half the length of the parking space;
and the processing module for inputting the expanded parking space image to the neural network to obtain at least one of area information and corner point information of an empty parking space from the parking space image. 14. The method according to any one of claims 10 to 13,
The device is
a second acquisition module for acquiring a plurality of parking space training images; and
Using the plurality of parking space training images, further comprising a training module for training the neural network,
The parking space training image is a parking space detection device including tagging information on key point information of an empty parking space. 15. The method of claim 14,
Extending the peripheral edge of the parking space training image outward by a preset value, wherein the preset value is less than or equal to half the length of the parking space. 16. The method of claim 14 or 15,
In the parking space training image, key point information of an empty parking space includes at least one corner point information of the empty parking space. 17. The method of claim 16,
the training module acquires area information constituted by corner point information of an empty parking space in the parking space training image and key point information of an empty parking space in the parking space training image; A parking space detection apparatus for training the neural network by using corner point information and area information of an empty parking space in the parking space training image and the parking space training image. 18. The method according to any one of claims 14 to 17,
The parking space training image is an image taken with a wide-angle camera. As an electronic device,
a memory in which computer programs are stored; and
An electronic device comprising a processor that executes the computer program to implement the method of detecting a parking space according to any one of claims 1 to 9. A computer storage medium storing a computer program, comprising:
A computer storage medium implementing the method of detecting a parking space according to any one of claims 1 to 9 when the computer program is executed.

Images