CN110501018B - Traffic sign information acquisition method for high-precision map production - Google Patents
Traffic sign information acquisition method for high-precision map production Download PDFInfo
- Publication number
- CN110501018B CN110501018B CN201910745743.4A CN201910745743A CN110501018B CN 110501018 B CN110501018 B CN 110501018B CN 201910745743 A CN201910745743 A CN 201910745743A CN 110501018 B CN110501018 B CN 110501018B
- Authority
- CN
- China
- Prior art keywords
- traffic sign
- information
- image
- position information
- traffic
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/26—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 specially adapted for navigation in a road network
- G01C21/28—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 specially adapted for navigation in a road network with correlation of data from several navigational instruments
- G01C21/30—Map- or contour-matching
- G01C21/32—Structuring or formatting of map data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
- G06V20/582—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads of traffic signs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
- G06V20/63—Scene text, e.g. street names
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/14—Image acquisition
- G06V30/146—Aligning or centring of the image pick-up or image-field
- G06V30/1475—Inclination or skew detection or correction of characters or of image to be recognised
- G06V30/1478—Inclination or skew detection or correction of characters or of image to be recognised of characters or characters lines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
Abstract
The invention relates to the field of high-precision map information acquisition, and discloses a traffic sign information acquisition method and system for high-precision map production, which comprises the steps of acquiring traffic scene image information to be identified; acquiring position information of the traffic sign board in the image; acquiring position information of the traffic sign in the image; the position information of the traffic sign in the image and the position information of the traffic sign in the image are used for attribution judgment, and the traffic sign information is attributed to the corresponding traffic sign information; the position information of the traffic sign board in the high-precision map is obtained, the traffic sign information and the corresponding traffic sign board information are added into the high-precision map information, so that the data storage is facilitated, the positioning of the traffic sign board is accurate, and the correct guidance is provided when the high-precision map information is used for navigation and the like.
Description
Technical Field
The invention relates to the field of high-precision map information acquisition, in particular to a traffic sign information acquisition method for serving high-precision map production.
Background
The high-precision map contains a large amount of driving assistance information, wherein traffic sign information is particularly important, the collection of the traffic sign information is used as an important module for producing the high-precision map, and most of the existing technical means are traffic sign detection combined with GPS positioning measurement.
At present, in the aspect of traffic sign board detection, a visual detection method is mainly adopted to obtain the position information of the traffic sign board, for example, a traffic sign detection and identification method based on a residual error SSD model with the patent number of CN108960198A, and multi-scale blocking is carried out on an image; and (3) constructing a residual SSD model by using a residual error network ResNet101 as a basic network of the SSD, carrying out network training, completing detection and identification with generalization capability, and realizing effective detection and identification of multiple types of signboards with different sizes in the real traffic scene of China. For example, a traffic sign detection method based on a visual attention mechanism and geometric features, which is disclosed in patent No. CN108256467A, organically combines the visual attention mechanism with the geometric features of a traffic sign according to the characteristics of the traffic sign, improves the conventional visual attention mechanism, introduces geometric feature constraints of the traffic sign, eliminates interference, and realizes distance measurement of the traffic sign.
In the conventional scheme, only the detection and recognition of the traffic sign in the image or the study of the distance measurement and positioning of the traffic sign are involved, and the high-precision map information requires more detailed traffic sign information, for example, information for guiding the user to travel, such as turn information and speed limit information.
Disclosure of Invention
In view of the problems faced by the background art, the present invention aims to provide a simple method and system for collecting traffic sign information with detailed traffic sign information, which is used for high-precision map production.
In order to achieve the purpose, the invention adopts the following technical scheme: a traffic sign information acquisition method for serving high-precision map production comprises the following steps: acquiring traffic scene image information to be identified; acquiring position information of the traffic sign board in the image; acquiring position information of the traffic sign in the image; the position information of the traffic sign in the image and the position information of the traffic sign board in the image are used for attribution judgment, and the traffic sign information is related to the corresponding traffic sign board information; and acquiring the position information of the traffic sign in the high-precision map, and adding the traffic sign information and the corresponding traffic sign information into the high-precision map information.
Preferably, the traffic sign and the traffic sign in the image are detected simultaneously by adopting an anchor-free deep learning target detection method, and the position information of the parallel minimum circumscribed rectangle frame of the traffic sign in the image and the position information of the parallel minimum circumscribed rectangle of the traffic sign in the image are respectively obtained.
Preferably, the distance between the positioning terminal and the traffic sign board is acquired by utilizing the position information of the parallel minimum circumscribed rectangular frame of the traffic sign board in the image; acquiring the position information of a positioning terminal by combining a real-time dynamic carrier phase differential positioning technology and an inertia measurement unit; and acquiring the position information of the traffic sign board in a high-precision map by combining the position information of the positioning terminal and the distance information between the positioning terminal and the traffic sign board.
Preferably, the position information of four corners of the traffic sign board is acquired, the actual distance of pixels in a rectangle formed by connecting the positioning terminal and the four corners is calculated by a binocular ranging method, and the average distance of the actual distances is used as the distance between the positioning terminal and the traffic sign board.
Preferably, the parallel minimum external rectangular frame is outwards expanded to be a rectangular enlarged frame around the central point of the rectangular enlarged frame, and the position information of the four corner points of the traffic sign board is obtained through key point detection in the rectangular enlarged frame.
Preferably, the image is corrected through the position information of four corner points of the traffic sign board; performing OCR recognition on the corrected image, and detecting auxiliary character information; and associating the auxiliary text information to corresponding traffic sign information, and adding the auxiliary text information into the high-precision map information.
Preferably, the position information of the parallel minimum circumscribed rectangle frame of the traffic sign in the image comprises center coordinate data, width data and height data of the traffic sign; the position information of the parallel minimum circumscribed rectangle of the traffic sign in the image comprises the center coordinate data of the traffic sign; and comparing the center coordinate data of the traffic sign with the center coordinate data, the width data and the height data of the traffic sign board, and attributing the traffic sign information to the corresponding traffic sign board information.
Preferably, the category information of a large category of the traffic signs is acquired, the large category comprises a prompt sign, a warning sign and a prohibition sign, the category information of the large category is used as a target of the same category for detection, the detection result is used as the input of a traffic sign classifier, and the category information of a small category of the traffic signs is acquired, wherein the small category comprises speed limit, height limit and stop prohibition.
Preferably, a computer-readable storage medium has stored thereon a computer program which, when being executed by a processor, carries out the steps of the method of any of the above.
Preferably, a traffic sign information collecting system serving high-precision map production includes: the image acquisition module is used for acquiring the image information of the traffic scene to be identified; the traffic sign acquisition module is used for acquiring the position information of the traffic sign in the image; the traffic sign acquisition module is used for acquiring the position information of the traffic sign in the image; the attribution module is used for judging attribution by utilizing the position information of the traffic sign in the image and the position information of the traffic sign board in the image and associating the traffic sign information with the corresponding traffic sign board information; and the positioning module is used for acquiring the position information of the traffic sign board in the high-precision map and adding the traffic sign information and the corresponding traffic sign board information into the high-precision map information.
Compared with the prior art, the invention provides a traffic sign information acquisition method for serving high-precision map production, which is characterized by comprising the following steps: acquiring traffic scene image information to be identified; acquiring position information of the traffic sign board in the image; acquiring position information of the traffic sign in the image; the position information of the traffic sign in the image and the position information of the traffic sign board in the image are used for attribution judgment, and the traffic sign information is related to the corresponding traffic sign board information; and acquiring the position information of the traffic sign in the high-precision map, and adding the traffic sign information and the corresponding traffic sign information into the high-precision map information. After the traffic sign information is associated with the corresponding traffic sign information, the traffic sign information and the corresponding traffic sign information are added into the high-precision map information, so that the data storage is convenient, the positioning of the traffic sign is accurate, and the correct guidance is provided when the high-precision map information is used for navigation and the like, so that the traffic signs belonging to the same traffic sign appear at the same position on the high-precision map and are consistent with the position of the actual traffic sign appearing on the actual road, the correct prompt is convenient to give, and the attractiveness of the high-precision map is improved.
Drawings
Fig. 1 is a schematic flow chart of a traffic sign information collection method for high-precision map production according to an embodiment of the present invention;
FIG. 2 is a block diagram of a traffic sign information collection system of the present invention that serves high precision map production;
with the foregoing drawings in mind, certain embodiments of the disclosure have been shown and described in more detail below. These drawings and written description are not intended to limit the scope of the disclosed concepts in any way, but rather to illustrate the concepts of the disclosure to those skilled in the art by reference to specific embodiments.
Detailed Description
Reference will now be made in detail to the exemplary embodiments, some of which are illustrated in the accompanying drawings and described below, wherein like reference numerals refer to like elements throughout. All other embodiments, which can be obtained by a person skilled in the art without any inventive step, based on the embodiments and the graphics of the invention, are within the scope of protection of the invention.
The application scenario of the embodiment of the invention is explained as follows, each intelligent navigation terminal realizes high-precision navigation and automatically drives the automobile, and the like, the demand of the electronic high-precision map is increasing, the high-precision map comprises a large amount of driving assistance information, such as current lane information, traffic sign information, traffic light information and the like, the embodiment of the invention provides a traffic sign information acquisition method for serving high-precision map production, wherein the traffic sign information is particularly important, so that in order to acquire high-precision map information with various detailed traffic sign information, the method can be applied to various acquisition terminal systems, wherein the acquisition terminal systems comprise a CCD camera system and a data processing system, and the data processing system is mainly used for storing and processing the shot related road data, traffic sign information and traffic sign information. The method may also be applied to other processing terminal systems including the data processing module, and the embodiment of the present invention is not particularly limited thereto. The acquisition terminal system can perform related operations by taking an automobile as a carrier.
The CCD camera in this embodiment is binocular, and the CCD camera in other embodiments may be three or more, again without limitation. The CCD camera may photograph the area containing the traffic sign to obtain image information containing the traffic sign, which the data processing system may store and process.
The acquisition terminal system not only can comprise a dual-purpose CCD camera system and a data processing system, but also can comprise other hardware and systems. Such as a processor, which is not limited in this embodiment of the present invention. The processor may be coupled to the CCD camera and the processor may perform correlation processing on the image information including the traffic sign.
The acquisition terminal system not only can comprise a CCD camera system and a data processing system, but also can comprise other systems, such as a data transmission system, and the data transmission system can transmit the acquired high-precision map information containing traffic sign information and traffic sign board information to a cloud end or a server. Therefore, high-precision navigation of the intelligent terminal or automatic driving of automobiles and the like is realized, and the embodiment of the invention is not particularly limited.
It should be further noted that, in the embodiment of the present invention, the collecting terminal system may obtain the image including the traffic sign and the traffic sign board through the CCD camera, and in other embodiments, the collecting terminal may also obtain the image including the traffic sign and the traffic sign board through other manners.
In the embodiment of the invention, the acquisition terminal system can measure the distance by using a CCD camera binocular vision distance measurement method, a three-eye distance measurement method and the like so as to improve the accuracy of distance measurement.
Fig. 1 is a schematic flow chart of a traffic sign information collection method for high-precision map production according to an embodiment of the present invention. The method comprises the following steps:
and S1, acquiring the image information of the traffic scene to be identified.
The traffic sign board is a facility for transmitting specific information by using graphic symbols and characters to manage traffic and indicate driving direction so as to ensure smooth road and safe driving. The highway is suitable for highways, urban roads and all special highways, has the property of law, and vehicles and pedestrians must obey the property. Because the surface color of the traffic sign board is gorgeous, the distinguishing capability of the traffic sign board is effectively improved, and a CCD camera can conveniently shoot clear images containing the traffic sign board.
Specifically, the collection terminal is provided with a CCD camera in each direction, the CCD camera shoots scenes in each direction to obtain traffic scene image information, some of the traffic scene images include a traffic sign image and a traffic sign image, some of the traffic scene images do not include the traffic sign image and the traffic sign image, and it should be noted that the collection terminal can also obtain interest point information of restaurants, hotels, shopping malls, gas stations, parking lots, and the like around a road besides recording various attribute information of the road. The monitoring range for acquiring the traffic scene image is large, road information can be comprehensively collected, and the equipment cost is low.
And S2, acquiring the position information of the traffic sign in the image.
Specifically, the deep learning target detection method adopting anchor-free mainly comprises two parts of image coding and decoding of the image, wherein the image coding adopts VGG16 as a backbone part, and 3x3 small convolution kernels and 2x2 maximum pooling layers are repeatedly stacked, so that the method is favorable for acquiring semantic information in a perception field. And the image decoding adopts an up-sampling method to amplify the coding part, which realizes that 2 times of deconvolution operation is continuously carried out, and adopts a convolution kernel of 1x1 to carry out regression prediction on the basis of the decoding characteristics, so as to directly obtain the position information of the traffic sign in the image, and the obtained result is the minimum circumscribed rectangle of each traffic sign in the image.
Specifically, the first step: the network input size is fixed, and after a detection image is obtained, the image is zoomed into a fixed size 640x352x 3;
step two: inputting the image into a VGG16 network, carrying out forward reasoning calculation, and carrying out a large amount of convolution and pooling operations;
step three: after the forward reasoning calculation is finished, obtaining coded image characteristics which are 1/32 of the original image and have the size of 20x11x 512;
step four: carrying out 8 times of up-sampling decoding operation on the coding characteristics, specifically continuous 2x2 deconvolution operation;
step five: after upsampling, obtaining the characteristics of a decoded image, wherein the size of the characteristics is 160x88x 128;
step six: performing regression prediction on the decoding characteristics by using a convolution kernel of 1x1 to obtain category prediction and box prediction information;
step seven: and combining the category prediction and the box prediction information to extract the box information predicted to be the traffic sign.
And S3, acquiring the position information of the traffic sign in the image.
Specifically, a traffic sign is a road facility that transmits guidance, restriction, warning, or indication information using words or symbols. Also known as road signs, road traffic signs. In traffic signs, traffic management is generally implemented by safe, eye-catching, clear and bright traffic signs, and important measures for ensuring the safety and smoothness of road traffic are taken. The traffic signs mainly include a prompt sign, a warning sign, a prohibition sign and other broad categories. The prompting marks play a role in indication, and are 29 categories in total, and the colors of the prompting marks are blue background and white patterns, such as straight running, turning left, rotary island running and the like. The warning signs mainly play a warning role, and are 49 subclasses in total, and the colors are yellow bottom, black edge and black patterns, such as cross, continuous turning, steep slope ascending and the like. The prohibition mark plays a role of prohibiting a certain behavior, and has a total of 43 categories, most of which are white background, red circle, red bar and black pattern and are arranged near the vehicle or the like which needs to be prohibited or is at present. For example, no passage, no entry, etc.
Specifically, the method for detecting the deep learning target by using the anchor-free method mainly comprises two parts of image coding and decoding of the image, and the method for acquiring the position information of the traffic sign in the image is consistent with the method for acquiring the position information of the traffic sign in the image, so that the position information of the traffic sign in the image is acquired, and the acquired result is the minimum circumscribed rectangle of each traffic sign in the image.
And S4, performing attribution judgment by using the position information of the traffic sign in the image and the position information of the traffic sign in the image, and associating the traffic sign information with the corresponding traffic sign information.
And comparing the center coordinate data of the traffic sign with the center coordinate data, the width data and the height data of the traffic sign board, and attributing the traffic sign information to the corresponding traffic sign board information.
One traffic sign board comprises a plurality of traffic signs, specifically, whether the center coordinates of the detected traffic signs are within the circumscribed rectangle range of the traffic sign board is judged, if yes, the traffic signs are judged to belong to the traffic sign board, and corresponding traffic sign information is related to the corresponding traffic sign board information, and the process mainly comprises the following steps:
the method comprises the following steps: acquiring position information of the traffic sign in the image by adopting an anchor-free target detection method, supposing that the obtained central coordinate of the traffic sign is A [ x1, y1], and the width and the height are W and H respectively;
step two: acquiring position information of a traffic sign in an image by adopting an anchor-free target detection method, supposing that the position information is a prompt traffic sign, the central coordinate of the prompt traffic sign is B [ x2, y2], and the width and the height are W2 and H2 respectively;
step three: inputting the images within the range of the prompting traffic sign into a classification network, such as alexnet, and further classifying, wherein the obtained classification is a right turn sign;
step four: judging whether x2 is in the range of [ x1-W/2, x1+ W/2] or not, whether y2 is in the range of [ y1-H/2, xy1+ H/2] or not, if yes, judging that the coordinate of the right-turn traffic sign attribution center is the traffic sign board with the A coordinate, and if not, judging one traffic sign board until the corresponding traffic sign board is matched or traversal is finished;
step five: and if the traffic sign is judged to belong to the corresponding traffic sign board, the traffic sign information is related to the corresponding traffic sign board information.
And S5, acquiring the position information of the traffic sign in the high-precision map, and adding the traffic sign information and the corresponding traffic sign information into the high-precision map information.
Specifically, the distance between the acquisition terminal and a pixel near the central point of the traffic sign is acquired by a binocular ranging method, the position information of the acquisition terminal is acquired by an RTK technology, and the position information of the traffic sign in a high-precision map is obtained after the position information and the position information are added. And finally, the traffic sign information and the corresponding traffic sign information are added into the high-precision map information to obtain the high-precision map information with detailed traffic sign information, so that the application in navigation or the automatic driving application is facilitated, the traffic sign information is associated into the corresponding traffic sign information, and then the traffic sign information and the corresponding traffic sign information are added into the high-precision map information, so that the data storage is facilitated, the positioning of the traffic sign is accurate, and the correct guidance is provided when the high-precision map information is used for navigation and other applications, so that the traffic signs belonging to the same traffic sign appear at the same position on the high-precision map and are consistent with the position of the actual traffic sign on the actual road, the correct prompt is facilitated, and the attractiveness of the high-precision map is improved.
The position information of the parallel minimum circumscribed rectangular frame of the traffic sign in the image comprises center coordinate data, width data and height data of the traffic sign; the position information of the parallel minimum circumscribed rectangle of the traffic sign in the image comprises the center coordinate data of the traffic sign;
specifically, the first step: obtaining the position information of the traffic sign board by adopting an anchor-free target detection method, wherein if the obtained central coordinates of the traffic sign board are A [ x1, y1], the width and the height are W and H respectively;
step two: since the obtained detection box is the minimum circumscribed rectangle of the traffic sign, the inside of the circumscribed rectangle may contain some other background information, and if the part is also considered as a part of the traffic sign, a certain distance measurement error may be caused. Therefore, the range of the required distance measurement is reduced, the center of the distance measurement range is the center coordinate of the traffic sign board A [ x1, y1], but the width and the height are respectively W/10 and H/10, and the distance measurement range is reduced to a certain extent;
step three: measuring pixels in the range by adopting a binocular ranging method, and averaging the pixels to be used as the relative position of the current traffic sign board;
step four: and combining the RTK information to obtain the position information of the traffic sign in the high-precision map.
Fig. 2 is a block diagram of a traffic sign information collecting system for high-precision map production according to an embodiment of the present invention. The system comprises:
s10, an image acquisition module for acquiring the traffic scene image information to be identified;
s20, a traffic sign board obtaining module for obtaining the position information of the traffic sign board in the image;
s30, a traffic sign acquisition module for acquiring the position information of the traffic sign in the image;
s40, an attribution module for judging attribution by using the position information of the traffic sign in the image and associating the traffic sign information to the corresponding traffic sign information;
and S50, a positioning module for acquiring the position information of the traffic sign in the high-precision map and adding the traffic sign information and the corresponding traffic sign information into the high-precision map information.
It should be noted that the method for detecting the deep learning target of anchor-free is adopted to simultaneously detect the traffic sign and the traffic sign in the image, and respectively acquire the position information of the parallel minimum circumscribed rectangle frame of the traffic sign in the image and the position information of the parallel minimum circumscribed rectangle of the traffic sign in the image. Specifically, the image is detected by adopting an anchor-free deep learning target detection method, and the position of the traffic sign in the image, the category of the large category and the position of the traffic sign in the image can be directly obtained at the same time, namely the position information of the parallel minimum circumscribed rectangle frame of the traffic sign in the image and the position information of the parallel minimum circumscribed rectangle of the traffic sign in the image are obtained. The current commonly used data sets for detecting traffic signs include T100 and CCTSDB, which can classify the traffic signs at the same time of detection, and the data sets are classified into a large category, that is, generally classified into a prompt sign, a warning sign and a prohibition sign. The classification of the subclasses such as steering and straight traveling requires secondary classification.
It should be noted that, due to the problem of unbalanced data sample number, more specific classification, such as a speed limit height limit sign, cannot be directly performed. The embodiment of the invention carries out classification twice, wherein the classification is carried out for the first time into three major classes, and the detailed subclass classification is carried out for the second time. Specifically, the category information of a large category of the traffic signs is obtained, the large category comprises a prompt sign, a warning sign and a prohibition sign, the category information of the large category is used as a target of the same category for detection, the detection result is used as the input of a traffic sign classifier, and the category information of a small category of the traffic signs is obtained, wherein the small category comprises speed limit, height limit, stopping prohibition and the like. Because the classification in the target detection is easily influenced by the target background, and the image after single-target detection basically only comprises the target object and a small amount of background information, the influence of the background can be effectively reduced. And for the training sample of the classifier, the corresponding data enhancement can be more conveniently obtained and carried out, and the problem of data imbalance is reduced.
It should be noted that the traffic sign board includes a plurality of traffic signs, that is, a plurality of traffic signs are located on the same traffic sign board, and in the high-precision map, it is necessary to record the traffic sign information into the corresponding traffic sign board information. This facilitates storage of high-precision map data and the aesthetic appearance of high-precision maps. Moreover, if the traffic signs on one traffic sign board are respectively positioned, it is likely that the traffic signs on the same traffic sign board are located at several places with different positions on the high-precision map due to deviation, which results in misjudgment of the user on the road section guidance. Such scattered data is also inconvenient to store.
It should be noted that, the distance between the positioning terminal and the traffic sign board is obtained by using the position information of the parallel minimum circumscribed rectangular frame of the traffic sign board in the image; the positioning terminal is the acquisition terminal. The acquisition terminal is arranged on the acquisition equipment, the positioning terminal comprises a related positioning device required by data acquisition, such as an RTK module and an IMU module, and the position information of the positioning terminal is acquired by combining a real-time dynamic carrier phase differential positioning technology and an inertial measurement unit; the position information of the positioning terminal can still be accurately obtained under the condition of weak signals.
And acquiring the position information of the traffic sign board in a high-precision map by combining the position information of the positioning terminal and the distance information between the positioning terminal and the traffic sign board. Acquiring position information of four corners of the traffic sign board, calculating the actual distance of pixels in a rectangle formed by connecting lines of the positioning terminal and the four corners by a binocular ranging method, and taking the average distance of the actual distances as the distance between the positioning terminal and the traffic sign board. It should be noted that, because the parallel minimum bounding rectangle usually includes the traffic sign and other background information, it is easy to have a large influence on the final ranging result. On the basis of target detection, the embodiment of the invention obtains four corner points of the traffic signboard in the parallel minimum circumscribed rectangle frame through key point detection, such as mtcnn and the like, calculates the actual distance of pixels in a corner point connecting line, and takes the average distance as the distance between the current traffic signboard and the positioning terminal. And finally, combining the position information of the positioning terminal and the distance between the current traffic sign board and the positioning terminal to obtain world coordinate data of the current traffic sign board, and adding high-precision map information.
Specifically, the parallel minimum external rectangular frame is outwards expanded to be a rectangular increasing frame around the central point of the rectangular minimum external rectangular frame, and the position information of the four corner points of the traffic sign board is obtained through key point detection in the rectangular increasing frame. Because the parallel minimum external rectangular frame may be tightly attached to the periphery of the traffic sign in the image, the traffic sign is not right opposite due to the influence of the photographing angle and the like, and the minimum external rectangular frame may not be capable of surrounding four corner points of the traffic sign in the image in the frame, the minimum external rectangular frame is firstly outwards expanded around the central point of the minimum external rectangular frame into a rectangular enlarged frame, and then the position information of four corner points of the traffic sign is detected, but the rectangular enlarged frame usually comprises the traffic sign and other background information, so that the final distance measurement result is easily influenced greatly. Therefore, the actual distance between the pixels in the rectangular range and the positioning terminal obtained by connecting the four corner points is used as the distance of the current traffic sign board, and the average distance of the actual distances is used as the distance of the current traffic sign board.
Specifically, the image is corrected through the position information of four corner points of the traffic sign board; performing OCR recognition on the corrected image, and detecting auxiliary character information; and associating the auxiliary text information to corresponding traffic sign information, and adding the auxiliary text information into the high-precision map information. The method comprises the steps that the traffic sign board is provided with traffic sign information and auxiliary character information, the auxiliary character information occupies a large part of the whole traffic sign board, key point detection is carried out on the basis of the traffic sign board detection, image correction is carried out through the detected key points, and the corrected image is used as an input of an RCNN to carry out OCR recognition. And associating the corresponding text information to the corresponding traffic sign information, and adding the text information to the high-precision map information. The image is corrected and then detected, the accuracy rate of detection is convenient, because shooting angle problem etc., the characters may be inclined, after correcting, the characters are just to placing, the accuracy rate of detection is convenient for improve, the accuracy rate of high accuracy map that just also improves is convenient for give the user the correct guide, safe trip, promotion user experience.
The invention also discloses a computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of any of the methods described above.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above.
The various embodiments or features mentioned herein may be combined with each other as additional alternative embodiments without conflict, within the knowledge and ability level of those skilled in the art, and a limited number of alternative embodiments formed by a limited number of combinations of features not listed above are still within the scope of the present disclosure, as understood or inferred by those skilled in the art from the figures and above.
Finally, it is emphasized that the above-mentioned embodiments, which are typical and preferred embodiments of the present invention, are only used for explaining and explaining the technical solutions of the present invention in detail for the convenience of the reader, and are not used to limit the protection scope or application of the present invention.
Therefore, any modifications, equivalents, improvements and the like made within the spirit and principle of the present invention should be covered within the protection scope of the present invention.
Claims (10)
1. A traffic sign information acquisition method for serving high-precision map production is characterized by comprising the following steps:
acquiring traffic scene image information to be identified;
acquiring position information of the traffic sign board in the image;
acquiring position information of the traffic sign in the image;
judging whether the traffic sign belongs to the traffic sign board or not according to whether the central coordinate of the traffic sign is within the range of the circumscribed rectangle of the traffic sign board or not, if so, judging that the traffic sign belongs to the traffic sign board, and associating traffic sign information into corresponding traffic sign board information;
and acquiring the position information of the traffic sign in the high-precision map, and adding the traffic sign information, the corresponding traffic sign information and the position information of the corresponding traffic sign in the high-precision map into the high-precision map information.
2. The method of claim 1, wherein: the method for detecting the deep learning target of the anchor-free is adopted to simultaneously detect the traffic sign board and the traffic sign in the image, and respectively obtain the position information of the parallel minimum circumscribed rectangle frame of the traffic sign board in the image and the position information of the parallel minimum circumscribed rectangle of the traffic sign in the image.
3. The method of claim 2, wherein: acquiring the distance between a positioning terminal and a traffic sign board by utilizing the position information of the parallel minimum circumscribed rectangular frame of the traffic sign board in the image;
acquiring the position information of a positioning terminal by combining a real-time dynamic carrier phase differential positioning technology and an inertia measurement unit;
and acquiring the position information of the traffic sign board in a high-precision map by combining the position information of the positioning terminal and the distance information between the positioning terminal and the traffic sign board.
4. The method of claim 3, wherein: acquiring position information of four corners of the traffic sign board, calculating the actual distance of pixels in a rectangle formed by connecting lines of the positioning terminal and the four corners by a binocular ranging method, and taking the average distance of the actual distances as the distance between the positioning terminal and the traffic sign board.
5. The method of claim 4, wherein:
and (3) outwards expanding the parallel minimum external rectangular frame around the central point of the minimum external rectangular frame into a rectangular enlarged frame, and detecting through key points in the rectangular enlarged frame to obtain the position information of four corner points of the traffic sign board.
6. The method of claim 4, wherein:
correcting the image through the position information of four corner points of the traffic sign board;
performing OCR recognition on the corrected image, and detecting auxiliary character information;
and associating the auxiliary text information to corresponding traffic sign information, and adding the auxiliary text information into the high-precision map information.
7. The method of claim 2, wherein:
the position information of the parallel minimum circumscribed rectangular frame of the traffic sign in the image comprises center coordinate data, width data and height data of the traffic sign;
the position information of the parallel minimum circumscribed rectangle of the traffic sign in the image comprises the center coordinate data of the traffic sign;
and comparing the center coordinate data of the traffic sign with the center coordinate data, the width data and the height data of the traffic sign board, and attributing the traffic sign information to the corresponding traffic sign board information.
8. The method of claim 1, wherein: the method comprises the steps of obtaining category information of a large category of the traffic signs, wherein the large category comprises a prompt sign, a warning sign and a prohibition sign, detecting the category information of the large category as a target of the same category, using a detection result as input of a traffic sign classifier, and obtaining the category information of a small category of the traffic signs, wherein the small category comprises speed limit, height limit and stop prohibition.
9. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the method according to any one of claims 1 to 8.
10. A traffic sign information collection system serving high-precision map production, comprising:
the image acquisition module is used for acquiring the image information of the traffic scene to be identified;
the traffic sign acquisition module is used for acquiring the position information of the traffic sign in the image;
the traffic sign acquisition module is used for acquiring the position information of the traffic sign in the image;
the attribution module is used for judging attribution by utilizing the position information of the traffic sign in the image and the position information of the traffic sign board in the image and associating the traffic sign information with the corresponding traffic sign board information;
and the positioning module is used for acquiring the position information of the traffic sign in the high-precision map and adding the traffic sign information, the corresponding traffic sign information and the position information of the corresponding traffic sign in the high-precision map into the high-precision map information.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910745743.4A CN110501018B (en) | 2019-08-13 | 2019-08-13 | Traffic sign information acquisition method for high-precision map production |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910745743.4A CN110501018B (en) | 2019-08-13 | 2019-08-13 | Traffic sign information acquisition method for high-precision map production |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110501018A CN110501018A (en) | 2019-11-26 |
| CN110501018B true CN110501018B (en) | 2021-11-02 |
Family
ID=68587509
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910745743.4A Active CN110501018B (en) | 2019-08-13 | 2019-08-13 | Traffic sign information acquisition method for high-precision map production |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110501018B (en) |
Families Citing this family (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112880693A (en) * | 2019-11-29 | 2021-06-01 | 北京市商汤科技开发有限公司 | Map generation method, positioning method, device, equipment and storage medium |
| CN113011212B (en) * | 2019-12-19 | 2024-04-05 | 北京四维图新科技股份有限公司 | Image recognition method and device and vehicle |
| CN111291660B (en) * | 2020-01-21 | 2022-08-12 | 天津大学 | Anchor-free traffic sign identification method based on void convolution |
| CN111275000B (en) * | 2020-02-18 | 2023-05-02 | 广州敏视数码科技有限公司 | Traffic sign board detection method based on historical positioning data |
| CN113673281B (en) * | 2020-05-14 | 2023-10-03 | 百度在线网络技术（北京）有限公司 | Speed limit information determining method, device, equipment and storage medium |
| CN112071078B (en) * | 2020-09-01 | 2022-11-08 | 交科院检测技术（北京）有限公司 | Traffic engineering environment intelligent detection system |
| CN112149624B (en) * | 2020-10-16 | 2022-06-10 | 腾讯科技（深圳）有限公司 | Traffic identification image processing method and device |
| CN112101299B (en) * | 2020-11-02 | 2021-02-19 | 立得空间信息技术股份有限公司 | Automatic traffic sign extraction method and system based on binocular CCD camera |
| CN112132853B (en) * | 2020-11-30 | 2021-02-05 | 湖北亿咖通科技有限公司 | Method and device for constructing ground guide arrow, electronic equipment and storage medium |
| CN112509329B (en) * | 2020-12-05 | 2021-10-08 | 武汉中海庭数据技术有限公司 | Traffic sign data processing method, electronic device and storage medium |
| CN113074749B (en) * | 2021-06-07 | 2021-08-20 | 湖北亿咖通科技有限公司 | Road condition detection and update method, electronic equipment and computer-readable storage medium |
| CN113628168A (en) * | 2021-07-14 | 2021-11-09 | 深圳海翼智新科技有限公司 | Target detection method and device |
| CN113947764B (en) * | 2021-12-06 | 2022-03-08 | 腾讯科技（深圳）有限公司 | Image processing method, device, equipment and storage medium |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN201269758Y (en) * | 2008-09-22 | 2009-07-08 | 交通部公路科学研究所 | Vehicle mounted full automatic detection recording system for traffic signs |
| CN106326858A (en) * | 2016-08-23 | 2017-01-11 | 北京航空航天大学 | Road traffic sign automatic identification and management system based on deep learning |
| CN106372577A (en) * | 2016-08-23 | 2017-02-01 | 北京航空航天大学 | Deep learning-based traffic sign automatic identifying and marking method |
| CN106525057A (en) * | 2016-10-26 | 2017-03-22 | 陈曦 | Generation system for high-precision road map |
| CN106980855A (en) * | 2017-04-01 | 2017-07-25 | 公安部交通管理科学研究所 | Traffic sign quickly recognizes alignment system and method |
| CN108090413A (en) * | 2017-11-21 | 2018-05-29 | 武汉中海庭数据技术有限公司 | A kind of traffic mark board correlating method and device |
| CN108230357A (en) * | 2017-10-25 | 2018-06-29 | 北京市商汤科技开发有限公司 | Critical point detection method, apparatus, storage medium, computer program and electronic equipment |
| CN109255279A (en) * | 2017-07-13 | 2019-01-22 | 深圳市凯立德科技股份有限公司 | A kind of method and system of road traffic sign detection identification |
| CN109409366A (en) * | 2018-10-30 | 2019-03-01 | 四川长虹电器股份有限公司 | Distorted image correction method and device based on Corner Detection |
Family Cites Families (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8509526B2 (en) * | 2010-04-13 | 2013-08-13 | International Business Machines Corporation | Detection of objects in digital images |
| JP5640794B2 (en) * | 2011-02-15 | 2014-12-17 | 株式会社Ｊｖｃケンウッド | Navigation device, navigation method, and program |
| CN104637300B (en) * | 2015-02-06 | 2017-01-04 | 南京理工大学 | Highway traffic sign board informationization Analysis of check-up and display systems |
| CN105718860B (en) * | 2016-01-15 | 2019-09-10 | 武汉光庭科技有限公司 | Localization method and system based on driving safety map and binocular Traffic Sign Recognition |
| JP6552979B2 (en) * | 2016-02-16 | 2019-07-31 | 株式会社日立製作所 | Image processing device, warning device, image processing system, and image processing method |
| US10859395B2 (en) * | 2016-12-30 | 2020-12-08 | DeepMap Inc. | Lane line creation for high definition maps for autonomous vehicles |
| CN107679508A (en) * | 2017-10-17 | 2018-02-09 | 广州汽车集团股份有限公司 | Road traffic sign detection recognition methods, apparatus and system |
| CN107957266B (en) * | 2017-11-16 | 2020-09-01 | 北京小米移动软件有限公司 | Positioning method, positioning device and storage medium |
| CN108846333B (en) * | 2018-05-30 | 2022-02-18 | 厦门大学 | Method for generating landmark data set of signpost and positioning vehicle |
-
2019
- 2019-08-13 CN CN201910745743.4A patent/CN110501018B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN201269758Y (en) * | 2008-09-22 | 2009-07-08 | 交通部公路科学研究所 | Vehicle mounted full automatic detection recording system for traffic signs |
| CN106326858A (en) * | 2016-08-23 | 2017-01-11 | 北京航空航天大学 | Road traffic sign automatic identification and management system based on deep learning |
| CN106372577A (en) * | 2016-08-23 | 2017-02-01 | 北京航空航天大学 | Deep learning-based traffic sign automatic identifying and marking method |
| CN106525057A (en) * | 2016-10-26 | 2017-03-22 | 陈曦 | Generation system for high-precision road map |
| CN106980855A (en) * | 2017-04-01 | 2017-07-25 | 公安部交通管理科学研究所 | Traffic sign quickly recognizes alignment system and method |
| CN109255279A (en) * | 2017-07-13 | 2019-01-22 | 深圳市凯立德科技股份有限公司 | A kind of method and system of road traffic sign detection identification |
| CN108230357A (en) * | 2017-10-25 | 2018-06-29 | 北京市商汤科技开发有限公司 | Critical point detection method, apparatus, storage medium, computer program and electronic equipment |
| CN108090413A (en) * | 2017-11-21 | 2018-05-29 | 武汉中海庭数据技术有限公司 | A kind of traffic mark board correlating method and device |
| CN109409366A (en) * | 2018-10-30 | 2019-03-01 | 四川长虹电器股份有限公司 | Distorted image correction method and device based on Corner Detection |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110501018A (en) | 2019-11-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110501018B (en) | Traffic sign information acquisition method for high-precision map production | |
| CN106919915B (en) | Map road marking and road quality acquisition device and method based on ADAS system | |
| CN110443225B (en) | Virtual and real lane line identification method and device based on feature pixel statistics | |
| CN101929867B (en) | Clear path detection using road model | |
| US8452103B2 (en) | Scene matching reference data generation system and position measurement system | |
| WO2018145602A1 (en) | Lane determination method, device and storage medium | |
| CN102682292B (en) | Method based on monocular vision for detecting and roughly positioning edge of road | |
| Guo et al. | Robust road detection and tracking in challenging scenarios based on Markov random fields with unsupervised learning | |
| Toulminet et al. | Vehicle detection by means of stereo vision-based obstacles features extraction and monocular pattern analysis | |
| CN101950350B (en) | Clear path detection using a hierachical approach | |
| CN104246821B (en) | Three-dimensional body detection device and three-dimensional body detection method | |
| Kong et al. | Generalizing Laplacian of Gaussian filters for vanishing-point detection | |
| Kühnl et al. | Spatial ray features for real-time ego-lane extraction | |
| CN109871776B (en) | All-weather lane line deviation early warning method | |
| EP2372607A2 (en) | Scene matching reference data generation system and position measurement system | |
| CN111563469A (en) | Method and device for identifying irregular parking behaviors | |
| CN105160309A (en) | Three-lane detection method based on image morphological segmentation and region growing | |
| CN101944176A (en) | Exist the more excellent clear path of means of transportation sign to detect | |
| CN110929655B (en) | Lane line identification method in driving process, terminal device and storage medium | |
| CN115717894B (en) | Vehicle high-precision positioning method based on GPS and common navigation map | |
| CN109635737A (en) | Automobile navigation localization method is assisted based on pavement marker line visual identity | |
| CN103770704A (en) | System and method for recognizing parking space line markings for vehicle | |
| CN107891808A (en) | Driving based reminding method, device and vehicle | |
| JP4761156B2 (en) | Feature position recognition apparatus and feature position recognition method | |
| CN106767854A (en) | mobile device, garage map forming method and system |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |