CN114694117A - Parking space identification method based on image - Google Patents
Parking space identification method based on image Download PDFInfo
- Publication number
- CN114694117A CN114694117A CN202210329810.6A CN202210329810A CN114694117A CN 114694117 A CN114694117 A CN 114694117A CN 202210329810 A CN202210329810 A CN 202210329810A CN 114694117 A CN114694117 A CN 114694117A
- Authority
- CN
- China
- Prior art keywords
- parking space
- line
- midpoint
- angular
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Traffic Control Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention provides an image-based parking space identification method, which adopts a convolutional neural network technology to extract angular point coordinates, parking space separation line directions and parking space warehousing line midpoint coordinates of parking spaces as parking space characteristics, and deduces the parking spaces through the characteristics, thereby realizing the function of detecting the parking space characteristics. The invention trains the neural network model by collecting tens of thousands of all-round images, and compared with the traditional image processing technology, the convolutional neural network technology greatly improves the identification accuracy of parking space detection. The method is suitable for more environmental scenes and is widely applied to the existing automatic parking space recognition technology.
Description
Technical Field
The invention belongs to the technical field of automobile electronic software, and particularly relates to a parking space identification method based on images.
Background
At present, intelligent automobiles are increasingly equipped with automatic parking technology. The automatic parking technology is to identify a parking space around a vehicle by using an on-board sensor and then control the vehicle to automatically park in the parking space. Parking stall discernment mainly has two kinds of modes: firstly, spatial parking space identification based on ultrasonic waves: the method mainly utilizes an ultrasonic sensor to identify a spatial parking space, and when the distance detected by the ultrasonic sensor jumps, the spatial parking space is indicated, and the method has the defects that reference vehicles are required to be arranged in front of and behind the parking space; secondly, marking-off parking space recognition based on a look-around camera: the method mainly comprises the steps of recognizing parking space characteristics such as parking space lines or parking space angular points in a look-around image by utilizing a computer vision technology, and then reasoning according to the parking space characteristics to obtain lineation parking spaces. The conventional image processing technology for identifying parking space features, such as identifying corners in an image by using a CANNY operator of OPENCV and identifying line segments in the image by using Hough transform, has the main disadvantages of low accuracy and few applicable scenes.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: the parking space recognition method based on the image is used for detecting parking space characteristics.
The technical scheme adopted by the invention for solving the technical problems is as follows: an image-based parking space recognition method comprises the following steps:
s1: training a convolutional neural network model for parking space feature recognition by using the look-around image, and extracting the corner coordinates of the parking space of the look-around image, the direction of the separation line of the parking space and the midpoint coordinates of the storage line of the parking space to obtain a convolutional neural network model weight for parking space feature recognition;
s2: detecting an angular point on the image to be detected, an angle of a parking space dividing line and a midpoint of a parking space warehousing line by using the trained model;
s3: deducing parking spaces according to the identified parking space angular point coordinates, the parting line angles and the midpoint coordinates of the parking space warehousing lines; the method comprises the step of reasoning the parking spaces based on two angular points and a midpoint of a parking space warehousing line or the step of reasoning the parking spaces based on one angular point and a midpoint of a parking space warehousing line.
According to the scheme, in the step S1, the specific steps are as follows:
the convolutional neural network model adopts Mobilene V1 as a backbone model, adopts a mean square error as the loss of the convolutional neural network model, and an Adam optimizer is used as the optimizer.
According to the scheme, in the step S1, the specific steps are as follows:
the identified parking space characteristics include: the parking space corner confidence level p1, the parking space corner coordinates x1 and y1, the directions cos a and sin a of the parking space separation line, the parking space inventory line midpoint confidence level p2, and the parking space inventory line midpoint coordinates x2 and y 2.
According to the scheme, in the step S3, the specific steps are as follows:
when two angular points and a center point of a warehousing line exist and the following three conditions are met, the two angular points are directly connected to infer a vertical parking space or an inclined parking space:
1) the distance between the two corner points is between [ L1-r 1], and L1+ r1 ];
2) the distance between the midpoint of the connecting line of the two corner points and the midpoint of the warehousing line is smaller than a threshold value r 1;
3) the angular difference between the two corner points is less than 10 degrees.
According to the scheme, in the step S3, the specific steps are as follows:
when two angular points and a center point of a warehousing line exist and the following three conditions are met, the two angular points are directly connected to deduce a horizontal parking space:
1) the distance between the two corner points is between [ L2-r2, L2+ r2 ];
2) the distance between the midpoint of the connecting line of the two corner points and the midpoint of the warehousing line is smaller than a threshold value r 2;
3) the angular difference between the two corner points is less than 10 degrees.
According to the scheme, in the step S3, the specific steps are as follows:
when an angular point and a center point of an entry line exist and the following two conditions are met, another angular point coordinate is calculated according to the angular point and the center point of the entry line, and the two angular points are directly connected to form a vertical parking space or an inclined parking space:
1) the distance between the corner point and the center point of the warehousing line is 0.5 x [ L1-r1, L1+ r1 ];
2) the distance between the midpoint of the connecting line of the two corner points and the midpoint of the warehousing line is less than the threshold value of 0.5 × r 1.
According to the scheme, in the step S3, the specific steps are as follows:
when one angular point and one warehousing line central point exist and the following two conditions are met, calculating another angular point coordinate according to the angular point and the warehousing line central point, wherein the two angular points are directly connected to form a horizontal parking space:
1) the distance between the corner point and the center point of the warehousing line is 0.5 x [ L2-r2, L2+ r2 ];
2) the distance between the midpoint of the connecting line of the two corner points and the midpoint of the warehousing line is less than the threshold value of 0.5 × r 2.
A computer storage medium having stored therein a computer program executable by a computer processor, the computer program performing an image-based parking space recognition method.
The invention has the beneficial effects that:
1. according to the image-based parking space identification method, the convolutional neural network technology is adopted to extract the angular point coordinates of the parking space, the direction of the parking space separation line and the midpoint coordinates of the parking space warehousing line to serve as the parking space characteristics, the parking space is deduced through the characteristics, and the function of detecting the parking space characteristics is achieved.
2. The invention trains the neural network model by collecting tens of thousands of all-round images, and compared with the traditional image processing technology, the convolutional neural network technology greatly improves the identification accuracy of parking space detection.
3. The method is suitable for more environmental scenes and is widely applied to the existing automatic parking space recognition technology.
Drawings
FIG. 1 is a flow chart of an embodiment of the present invention.
FIG. 2 is a diagram of an input image and a feature matrix according to an embodiment of the invention.
Fig. 3 is a parking space characteristic diagram according to an embodiment of the present invention.
Fig. 4 is a first neural network model weight diagram according to an embodiment of the present invention.
Fig. 5 is a second neural network model weight graph according to an embodiment of the present invention.
Fig. 6 is a third diagram of the neural network model weights according to the embodiment of the present invention.
Fig. 7 is a fourth graph of neural network model weights according to an embodiment of the present invention.
FIG. 8 is a first inference diagram of an embodiment of the invention.
Fig. 9 is a second inference diagram of an embodiment of the invention.
FIG. 10 is a reasoning diagram III of an embodiment of the present invention.
Fig. 11 is a reasoning diagram four of an embodiment of the invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and specific embodiments.
Referring to fig. 1, a parking space recognition method based on an image according to an embodiment of the present invention includes the following steps:
s1: training is used for the neural network model of parking stall feature recognition, and the characteristic of discernment includes: a parking space corner confidence p1, parking space corner coordinates x1 and y1, directions cos a and sin a of a parking space separation line, a parking space inventory line midpoint confidence p2, and parking space inventory line midpoint coordinates x2 and y 2;
s2: detecting an angular point on the image, an angle of a parking space dividing line and a midpoint of a parking space warehousing line by using the trained model;
s3: deducing the parking space based on two angular points and the midpoint of a parking space storage line, and when the distance between the two angular points is [ L1-r 1], L1+ r1] in the condition of 1) is met; 2) the distance between the midpoint of the connecting line of the two corner points and the midpoint of the warehousing line is smaller than a threshold value r 1; 3) and when the angle difference value of the two angular points is smaller than three conditions of 10 degrees, the two angular points are directly connected to deduce a vertical parking space or an inclined parking space. When 1) the distance between two corner points is between [ L2-r2, L2+ r2 ]; 2) the distance between the midpoint of the connecting line of the two corner points and the midpoint of the warehousing line is smaller than a threshold value r 2; 3) and when the angle difference value of the two angular points is less than three conditions of 10 degrees, the two angular points are directly connected to deduce a horizontal parking space.
S4: deducing parking spaces based on one angular point and the midpoint of a parking space warehousing line, wherein when the distance between the angular point and the center point of the warehousing line meets 1) is 0.5 × L1-r1, L1+ r 1; 2) and when the distance between the midpoint of the connecting line of the two angular points and the midpoint of the warehousing line is smaller than two conditions of 0.5 × r1, calculating another angular point coordinate according to the angular point and the center point of the warehousing line, and directly connecting the two angular points to form a vertical parking space or an inclined parking space. When 1) the distance between the corner point and the center point of the warehousing line is 0.5 x [ L2-r2, L2+ r2 ]; 2) and when the distance between the midpoint of the connecting line of the two angular points and the midpoint of the warehousing line is less than two conditions of 0.5 × r2, calculating the coordinate of the other angular point according to the angular point and the center point of the warehousing line, and directly connecting the two angular points to form a horizontal parking space.
Referring to fig. 2, with the circled stitched image as input, the image input size is 512 × 512, and with mobilene V1 as the backbone model, the input image has a length and width of 16 after the multi-layer convolution operation in mobilene V1.
Referring to fig. 3, the number of channels of the design matrix is 8, which respectively represents the parking space corner confidence p1, the parking space corner coordinates x1 and y1, the directions cos a and sin a of the parking space partition line, the parking space inventory line midpoint confidence p2, and the parking space inventory line midpoint coordinates x2 and y 2.
The loss of the convolutional neural network uses the mean square error and the optimizer uses an Adam optimizer. And obtaining the weight of the neural network model for identifying the parking space characteristics after enough training steps. See fig. 4, 5, 6, 7 for example results:
and deducing the parking space according to the identified parking space angular point coordinates, the parting line angles and the midpoint coordinates of the parking space warehousing line.
Referring to fig. 8, when there are two corner points and one warehousing line center point, if the following condition is satisfied:
(1) the distance between the two corner points is between [ L1-r 1], and L1+ r1 ];
(2) the distance between the midpoint of the connecting line of the two corner points and the midpoint of the warehousing line is smaller than a threshold value r 1;
(3) the angle difference of the two corner points is less than 10 degrees;
the two angular points may constitute a vertical parking space or an oblique parking space.
Referring to fig. 9, when there are two corner points and one warehousing line center point, if the following condition is satisfied:
(1) the distance between the two corner points is between [ L2-r 2], and L2+ r2 ];
(2) the distance between the midpoint of the connecting line of the two corner points and the midpoint of the warehousing line is smaller than a threshold value r 2;
(3) the angle difference of the two corner points is less than 10 degrees;
the two angular points may constitute a horizontal parking space.
Referring to fig. 10, when there is one corner point and one warehousing line center point, if the following condition is satisfied:
(1) the distance between the corner point and the center point of the warehousing line is 0.5 x [ L1-r1, L1+ r1 ];
(2) the distance between the midpoint of the connecting line of the two corner points and the midpoint of the warehousing line is less than 0.5 × r 1;
another angular point coordinate can be calculated according to the angular point and the center point of the warehousing line, and the two angular point coordinates can form a vertical parking space or an inclined parking space.
Referring to fig. 11, when there is one corner point and one warehousing line center point, if the following condition is satisfied:
(1) the distance between the corner point and the center point of the warehousing line is 0.5 x [ L2-r2, L2+ r2 ];
(2) the distance between the midpoint of the connecting line of the two corner points and the midpoint of the warehousing line is less than 0.5 × r 2;
another angular point coordinate can be calculated according to the angular point and the center point of the warehousing line, and the two angular point coordinates can form a horizontal parking space.
The above embodiments are only used for illustrating the design idea and features of the present invention, and the purpose of the present invention is to enable those skilled in the art to understand the content of the present invention and implement the present invention accordingly, and the protection scope of the present invention is not limited to the above embodiments. Therefore, all equivalent changes and modifications made in accordance with the principles and concepts disclosed herein are intended to be included within the scope of the present invention.
Claims (8)
1. An image-based parking space recognition method is characterized in that: the method comprises the following steps:
s1: training a convolutional neural network model for parking space feature recognition by using the look-around image, and extracting the corner coordinates of the parking space, the direction of the parking space partition line and the midpoint coordinates of the parking space warehousing line of the look-around image to obtain a convolutional neural network model weight for parking space feature recognition;
s2: detecting an angular point on the image to be detected, an angle of a parking space dividing line and a midpoint of a parking space warehousing line by using the trained model;
s3: deducing parking spaces according to the identified parking space angular point coordinates, the parting line angles and the midpoint coordinates of the parking space warehousing lines; the method comprises the step of reasoning the parking space based on two angular points and the midpoint of a parking space storage line or the step of reasoning the parking space based on one angular point and the midpoint of a parking space storage line.
2. The image-based parking space recognition method according to claim 1, wherein: in the step S1, the specific steps are as follows:
the convolutional neural network model adopts Mobilene V1 as a backbone model, adopts a mean square error as the loss of the convolutional neural network model, and an Adam optimizer is used as the optimizer.
3. The image-based parking space recognition method according to claim 1, wherein: in the step S1, the specific steps are as follows:
the identified parking space characteristics include: the parking space corner confidence level p1, the parking space corner coordinates x1 and y1, the directions cos a and sin a of the parking space separation line, the parking space inventory line midpoint confidence level p2, and the parking space inventory line midpoint coordinates x2 and y 2.
4. The image-based parking space recognition method according to claim 1, wherein: in the step S3, the specific steps are as follows:
when two angular points and a center point of a warehousing line exist and the following three conditions are met, the two angular points are directly connected to infer a vertical parking space or an inclined parking space:
1) the distance between the two corner points is between [ L1-r 1], and L1+ r1 ];
2) the distance between the midpoint of the connecting line of the two corner points and the midpoint of the warehousing line is smaller than a threshold value r 1;
3) the angular difference between the two corner points is less than 10 degrees.
5. The image-based parking space recognition method according to claim 1, wherein: in the step S3, the specific steps are:
when two angular points and a center point of a warehousing line exist and the following three conditions are met, the two angular points are directly connected to deduce a horizontal parking space:
1) the distance between the two corner points is between [ L2-r 2], and L2+ r2 ];
2) the distance between the midpoint of the connecting line of the two corner points and the midpoint of the warehousing line is smaller than a threshold value r 2;
3) the angular difference between the two corner points is less than 10 degrees.
6. The image-based parking space recognition method according to claim 1, wherein: in the step S3, the specific steps are:
when an angular point and a center point of an entry line exist and the following two conditions are met, another angular point coordinate is calculated according to the angular point and the center point of the entry line, and the two angular points are directly connected to form a vertical parking space or an inclined parking space:
1) the distance between the corner point and the center point of the warehousing line is 0.5 x [ L1-r1, L1+ r1 ];
2) the distance between the midpoint of the connecting line of the two corner points and the midpoint of the warehousing line is less than the threshold value of 0.5 × r 1.
7. The image-based parking space recognition method according to claim 1, wherein: in the step S3, the specific steps are as follows:
when one angular point and one warehousing line central point exist and the following two conditions are met, calculating another angular point coordinate according to the angular point and the warehousing line central point, wherein the two angular points are directly connected to form a horizontal parking space:
1) the distance between the corner point and the center point of the warehousing line is 0.5 x [ L2-r2, L2+ r2 ];
2) the distance between the midpoint of the connecting line of the two corner points and the midpoint of the warehousing line is less than the threshold value of 0.5 × r 2.
8. A computer storage medium, characterized in that: stored therein is a computer program executable by a computer processor, the computer program performing an image-based parking space recognition method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210329810.6A CN114694117A (en) | 2022-03-30 | 2022-03-30 | Parking space identification method based on image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210329810.6A CN114694117A (en) | 2022-03-30 | 2022-03-30 | Parking space identification method based on image |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114694117A true CN114694117A (en) | 2022-07-01 |
Family
ID=82141905
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210329810.6A Pending CN114694117A (en) | 2022-03-30 | 2022-03-30 | Parking space identification method based on image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114694117A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115297308A (en) * | 2022-07-29 | 2022-11-04 | 东风汽车集团股份有限公司 | Surrounding AR-HUD projection system and method based on unmanned aerial vehicle |
-
2022
- 2022-03-30 CN CN202210329810.6A patent/CN114694117A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115297308A (en) * | 2022-07-29 | 2022-11-04 | 东风汽车集团股份有限公司 | Surrounding AR-HUD projection system and method based on unmanned aerial vehicle |
| CN115297308B (en) * | 2022-07-29 | 2023-05-26 | 东风汽车集团股份有限公司 | Surrounding AR-HUD projection system and method based on unmanned aerial vehicle |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108537197B (en) | Lane line detection early warning device and method based on deep learning | |
| CN110246183B (en) | Wheel grounding point detection method, device and storage medium | |
| CN113874927A (en) | Parking detection method, system, processing device and storage medium | |
| CN111178236A (en) | Parking space detection method based on deep learning | |
| CN110889974B (en) | Intelligent parking space identification method and device and automobile | |
| CN112967283A (en) | Target identification method, system, equipment and storage medium based on binocular camera | |
| WO2020181426A1 (en) | Lane line detection method and device, mobile platform, and storage medium | |
| CN112509027B (en) | Repositioning method, robot, and computer-readable storage medium | |
| CN115717894A (en) | Vehicle high-precision positioning method based on GPS and common navigation map | |
| CN110689578A (en) | Unmanned aerial vehicle obstacle identification method based on monocular vision | |
| CN112198878B (en) | Instant map construction method and device, robot and storage medium | |
| CN114694117A (en) | Parking space identification method based on image | |
| CN110992424B (en) | Positioning method and system based on binocular vision | |
| CN111739333B (en) | Empty parking space identification method | |
| CN114550142A (en) | Parking space detection method based on fusion of 4D millimeter wave radar and image recognition | |
| CN112666573B (en) | Detection method for retaining wall and barrier behind mine unloading area vehicle | |
| CN111428616B (en) | Parking space detection method, device, equipment and storage medium | |
| CN116486130A (en) | Obstacle recognition method, device, self-mobile device and storage medium | |
| US20220335732A1 (en) | Method and system for recognizing surrounding driving environment based on svm original image | |
| CN114926820A (en) | Inclined parking space identification method and system based on deep learning and image frame optimization | |
| CN113536839B (en) | Data processing method and device, positioning method and device and intelligent equipment | |
| CN114821135A (en) | License plate position identification method and system and computer storage medium | |
| CN114783172B (en) | Parking lot empty space recognition method and system and computer readable storage medium | |
| US20230394843A1 (en) | Method for identifying moving vehicles and in-vehicle device | |
| CN118115975A (en) | Low-speed unmanned control method and system based on obstacle detection |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |