AUTOMATIC LICENSE PLATE READING METHOD IMPLEMENTED BY A LICENSE PLATE READING DEVICE

Abstract

The invention relates to a method for authenticating license plates implemented by an authentication device (2) comprising a camera (4) positioned opposite parking spaces (5): identification of the plates and acquisition of an image by the camera (4); transmission of said image to a processing unit (6) of the device (2); detection of a string of characters on each image; parallax correction on each image; reconstruction of the identified string of characters; storage of the strings of characters; the method being characterized in that the parallax correction applied is a function of an angle of incidence varying for the correction of each character. Abstract figure: Figure 1

Description

Title of the invention: METHOD FOR AUTOMATIC LICENSE PLATE READING IMPLEMENTED BY A DEVICE LICENSE PLATE READING technical field

[0001] The present invention relates to the field of vehicle presence detection equipment, associated with a system for counting and guiding users to their places in a parking lot, while taking into account the traffic flow of vehicles exiting or searching for parking spaces in said parking lot.

[0002] The present invention also relates to the field of vehicle authentication devices associated with such detection, counting and guidance equipment. STATE OF THE ART

[0003] The prior art of license plate recognition counting and authentication systems is known as described in US patent 6553131B1. These devices are generally known to those skilled in the art by the acronym "ALPR" for "Automatic Number Plate Recognition System." These devices use an optical character recognition process, known to those skilled in the art by the acronym "OCR." On average, the OCR character recognition rate for license plates is greater than 98% when acquisition devices such as cameras are correctly positioned to avoid errors. In a parking lot, these devices are advantageously positioned at checkpoints where vehicle authentication is required, for example, at a parking lot entrance.

[0004] It is also known from the prior art, notably from document WO2021144617, that an ANPR device, for example positioned at the entrance to a parking lot, and an acquisition device such as a camera facing a plurality of parking spaces can be used simultaneously. Such a device allows the recognition of license plates of vehicles parked on each side of a traffic lane, but has a significant limitation in the number of vehicles that can be authenticated from a single acquisition device. The number of license plates that can be recognized and authenticated is technically limited by the camera's field of view due to parallax error, i.e. This refers to the error related to image distortion due to changes in the viewing angle. When the camera identifies one or two license plates, known OCR methods are functional and exhibit virtually no error in authenticating the characters of the license plate(s). It has been observed that beyond two license plates visible from the camera's viewing angle, the angle of incidence of a plate relative to a camera reference axis is such that the recognition rate during authentication decreases significantly to approximately 80%. Such a rate is insufficient to correctly authenticate a vehicle, for example, due to confusion between two similar-looking characters, such as the Latin letter "O" and the number "0".It is therefore necessary to increase the number of cameras as the number of parking spaces increases, in order to reduce the viewing angles from each parking space, which necessarily represents a significant financial investment.

[0005] The invention aims to resolve all or part of the drawbacks of the prior art, by proposing an automatic license plate reading method that reduces the number of acquisition devices needed to cover all the parking spaces in a car park. PRESENTATION OF THE INVENTION

[0006] More specifically, the invention relates to a method for authenticating license plates implemented by a license plate authentication device, said device comprising at least one camera disposed on a traffic lane of a parking lot, opposite a plurality of parking spaces, such that the license plates of at least two vehicles parked in adjacent parking spaces are visible to the camera, each license plate comprising a first lateral end and a second lateral end, the first end of each plate forming a first angle with respect to a reference axis of the camera and the second end of each plate forming a second angle with respect to the reference axis of the camera, the first angle being greater than the second angle, the method comprising the following steps: • the identification of each of the license plates by the device's camera and the acquisition of an image of each of the license plates by the camera; • the transmission by the camera of said image of each of the license plates to a processing unit of the license plate authentication device; • the detection on each license plate image of an alphanumeric character string of the license plate; • parallax correction by the device's processing unit on each of the transmitted license plate images; • the reconstruction of the alphanumeric character string identified on each image by optical character recognition; • the storage in memory by the device's processing unit of the reconstructed alphanumeric character strings.

[0007] The method is remarkable in that the parallax correction applied to each of the license plate images is a function of an angle of incidence, the angle of incidence being for each license plate image between the first angle and the second angle with respect to the camera's reference axis, the angle of incidence varying for the correction of a parallax of each character detected on each license plate image.

[0008] The term “identification” means the act of detecting the presence of a license plate in the field of view of the camera of the device.

[0009] The term "authentication" means the process of creating a string of alphanumeric characters from an image of a license plate, the reconstructed string corresponding to the characters visible on said image.

[0010] The term "detection" means the process of individually delimiting each alphanumeric character on an image of a license plate, without proceeding to the authentication of said character by character recognition.

[0011] Thanks to this combination of features, such a method makes it possible to authenticate a greater number of vehicle license plates using the same acquisition equipment, particularly cameras. Furthermore, this method significantly reduces errors due to the parallax effect caused by the angle of incidence of the characters on a license plate with the camera, notably by applying a correction specific to each character, thus promoting superior optical character recognition.

[0012] Advantageously, the method includes a step of searching for each reconstructed alphanumeric string in a database of license plates acquired subsequently or previously. In such a configuration, each reconstructed string authenticating a given vehicle is compared to a string stored in a license plate database, making it possible, for example, to verify that the reconstructed string is indeed a valid and existing license plate number.

[0013] Advantageously, the database consists of alphanumeric character strings, each associated with an acquired license plate image by a camera on the device at one entrance to the parking lot, and / or by license plate images acquired by a camera on the device at one exit of the parking lot. In such a configuration, the search is performed on a small sample of license plates, limiting the number of possible errors.

[0014] Advantageously, if an alphanumeric string stored in memory corresponds to a license plate image from the database acquired by a camera on the device at the parking lot exit, said string is erased from memory. In such a configuration, the device only retains in memory those strings that could correspond to license plates of vehicles that have entered and are circulating in the parking lot.

[0015] Advantageously, the parallax correction applied includes an application of the Hough transform followed by a rotation of the image of each plate.

[0016] Advantageously, the parallax correction step is performed by an artificial intelligence of the processing unit, based on deep neural networks operating by learning.

[0017] According to another aspect of the invention, it relates to a license plate authentication device for a parking lot configured to implement the method described above, comprising: • at least one camera positioned on a traffic lane in the parking lot, opposite a plurality of parking spaces, so that the license plates of at least two vehicles parked in adjacent parking spaces are visible to the camera; • a processing unit connected to the camera and including a storage memory, the processing unit being configured to process images acquired by the camera; • the authentication device being configured to correct a parallax induced by an angle of incidence between a license plate and a reference axis of the camera.

[0018] Advantageously, the license plate authentication device comprises at least two cameras arranged on a traffic lane of the parking lot, each camera facing a plurality of parking spaces on a given side of the traffic lane. In such a configuration, the cameras make it possible to acquire images of license plates located on both sides of the traffic lane.

[0019] Advantageously, the license plate authentication device includes a camera positioned at an entrance to the parking lot and / or a camera positioned at an exit to the parking lot.

[0020] Advantageously, the processing unit includes an artificial intelligence based on deep neural networks operating by learning. PRESENTATION OF THE FIGURES

[0021] The invention will be better understood upon reading the following description, given solely by way of example, and referring to the accompanying drawings given by way of non-limiting examples, in which identical references are given to similar objects and on which:

[0022] Fig. 1 is a schematic representation of a parking lot comprising a plurality of parking spaces and equipped with a license plate authentication device according to a first embodiment of the invention;

[0023] Fig. 2 is a schematic representation of a plurality of parking spaces in relation to a camera of the authentication device;

[0024] Fig. 3 is a schematic representation of the detection of a license plate by the camera of the authentication device;

[0025] Fig. 4 is a schematic representation of license plate images according to a plurality of incidences with respect to a reference axis of the device's camera;

[0026] Fig. 5 is a schematic representation of a parallax correction and optical character recognition according to a step of an authentication process according to another aspect of the invention of a parallax correction step.

[0027] It should be noted that the figures set out the invention in detail to enable implementation of the invention; although not limiting, said figures serve in particular to better define the invention where appropriate. DETAILED DESCRIPTION OF THE INVENTION

[0028] Figure 1 represents a parking lot 1 comprising a plurality of parking spaces 5. The parking spaces 5 are configured to each accommodate a vehicle 3 belonging to a user wishing to park their vehicle 3. Each parking space 5 is adjacent to at least one other parking space 5, so that the parked vehicles 3 extend longitudinally and parallel to one another. Advantageously, pluralitys of parking spaces 5 extend on either side of a traffic lane so as to form rows of parking spaces 5.

[0029] According to a first aspect of the invention, the parking lot 1 is equipped with an authentication device 2 configured to authenticate the license plates 7 of the vehicles 3 in the parking lot 1. The authentication device 2 comprises at least one camera 4 disposed on the traffic lane of the parking lot 1, opposite one of the plurality of parking spaces 5. The camera 4 is positioned so that the The license plates 7 of at least two vehicles 3 parked in adjacent parking spaces 5 are visible to the camera 4. In other words, at least two license plates 7 are included within the field of view of the camera 4, in a horizontal plane parallel to the ground of the parking lot 1 as illustrated in [Fig. 2]. Said camera 4 has a reference axis A, preferably defined as the bisector of the field of view of the camera 4.

[0030] The authentication device 2 further comprises at least one processing unit 6. The processing unit 6 is configured to receive and process images 11 acquired by the camera 4. Where the device 2 comprises several cameras 4, each camera 4 is connected to the processing unit 6. The processing unit 6 may comprise subunits, each associated with a camera 4 and configured to perform local processing of the images 11 captured by the associated camera 4. In the embodiment described in [Fig. 1], the cameras 4 are positioned opposite the parking spaces 5 located on either side of the traffic lane of the parking lot 1.

[0031] The remainder of the description focuses on describing a method for authenticating license plates 7 implemented by the authentication device 2 according to another aspect of the invention. The term "authentication" refers to the process of constructing a string of alphanumeric characters 13 from an image 11 of a license plate 7, the reconstructed string corresponding to the characters visible on said image 11. Each license plate 7 comprises a first lateral end 9a and a second lateral end 9b. The first end 9a forms a first angle with respect to the reference axis A of the camera 4. The second end 9b forms a second angle with respect to the reference axis A of the camera 4, the first angle being greater than the second angle, as illustrated in [Fig. 3].

[0032] The method includes a step of identifying each of the license plates 7 by the camera 4 of the device 2 and acquiring an image 11 of each of the license plates 7 by the camera 4. The term "identification" means detecting the presence of a license plate 7 in the field of view of the camera 4 of the device 1. Naturally, the image 11 can correspond to either a snapshot or a video.

[0033] The method includes a step of transmitting by the camera 4 the image 11 of each of the license plates 7 to the processing unit 6 of the device 2. The transmission can for example be carried out by a cable connection in the case where the processing unit 6 is in situ, but can alternatively be carried out by wireless communication in the case where the processing unit 6 is located remotely, in particular via subunits as described above.

[0034] The method includes a detection step on each image 11 of an alphanumeric character string 13 of the license plate 7. By the term "Detection" refers to the process of individually delimiting each alphanumeric character 13 on an image 11 of a license plate 7, without authenticating said character. In other words, the camera 4 transmits information indicating the presence of a license plate 7 on the image 11, which allows, for example, the presence of a vehicle 3 in the associated parking space 5 to be indicated by a warning light. Detection may include, in particular, a calculation of Haar characteristics, advantageously utilizing the creation of full-frame images.

[0035] The method includes a parallax correction step performed by the processing unit 6 on each of the transmitted images 11. Naturally, the image 11 may require parallax correction, particularly when the license plate 7 is aligned with the reference axis A of the camera 4, as illustrated in [Fig. 3]. The correction applied to each image 11 is a function of an angle of incidence θ between the first and second angles with respect to the reference axis A of the camera 4, as illustrated in [Fig. 4]. The angle of incidence θ during the parallax correction of each character 13 detected on each image 11. Such a method makes it possible to authenticate a greater number of vehicle license plates 7 using the same number of cameras. Advantageously, the parallax correction applied includes an application of the Hough transform followed by a rotation of the image 11 of each plate 7.The image rotation is then based on angles obtained by the Hough transform and includes in particular the calculation of a 2D affine rotation matrix allowing to perform an affine transformation of an image 11 and the application of the affine transformation to said image 11. .

[0036] In particular, the method includes a step of reconstructing the alphanumeric character string identified on each image 11 by optical character recognition. Thanks to the correction step previously performed, such a method makes it possible to significantly reduce errors due to the parallax effect induced by the angle of incidence 0 on the characters 13, notably by applying a correction specific to each character 13, which promotes higher-quality optical character recognition, as illustrated in [Fig. 5]. Pre-processing of the image 11 can be carried out before optical character recognition, the pre-processing including, in particular, median filtering followed by a Canny filter.

[0037] The process includes in particular a step of storing in memory by the processing unit 6 the reconstituted alphanumeric character strings 13.

[0038] Advantageously, the method includes a step of searching for each reconstructed alphanumeric character string 13 in a license plate database 7. In such a configuration, each reconstructed character string 13 authenticating a given vehicle 3 is compared to a string of characters 13 stored in a license plate database, allowing for example to verify that the reconstructed string 13 is indeed a valid and existing license plate number 7.

[0039] The authentication device 2 may advantageously include a camera (not shown) located at an entrance 15 and / or a camera (not shown) at an exit 17 of the parking lot 1, connected to processing units 6 and enabling the reconstruction of alphanumeric character strings 13 corresponding to license plates 7 of vehicles 3 entering and / or exiting the parking lot 1. In such a configuration, the search is performed on a reduced sample of license plates 7, considerably increasing the recognition accuracy. The images 11 acquired by the cameras 4 at the entrance 15 and / or exit 17 of the parking lot 1 have satisfactory reliability, the license plates 7 of the vehicles 3 at these entry points being aligned with the reference axis A of the cameras 4 during the acquisition of the images 11.

[0040] Advantageously, the device 2, and more particularly the processing unit 6, comprises an artificial intelligence algorithm based on deep neural networks operating by learning. Preferably, the artificial intelligence algorithm is configured to implement the parallax correction step, thereby increasing the reliability of the correction over time.

[0041] It should also be noted that the invention is not limited to the embodiments described above. It will indeed be apparent to a person skilled in the art that various modifications can be made to the embodiment described above, in light of the information just disclosed to them.

[0042] In the detailed presentation of the invention given above, the terms used shall not be interpreted as limiting the invention to the embodiment set forth in this description, but shall be interpreted to include all equivalents which can be foreseen by a person skilled in the art by applying their general knowledge to the implementation of the teaching which has just been disclosed to them.

Claims

1. Claims A method for authenticating license plates (7) implemented by a device (2) for authenticating license plates (7), said device (2) comprising at least one camera (4) arranged on a traffic lane of a parking lot (2), opposite a plurality of parking spaces (5), so that license plates (7) of at least two vehicles (3) parked in adjacent parking spaces (5) are visible to the camera (4), each license plate (7) comprising a first lateral end (9a) and a second lateral end (9b), the first end (9a) of each plate (7) forming a first angle relative to a reference axis (A) of the camera (4) and the second end (9b) of each plate (7) forming a second angle relative to the reference axis (A) of the camera (4), the first angle being greater than the second angle, the method comprising the following steps: • the identification of each of the license plates (7) by the camera (4) of the device (2) and the acquisition of an image (11) of each of the license plates (7) by the camera (4); • the transmission by the camera (4) of said image (11) of each of the license plates (7) to a processing unit (6) of the device (2) for authenticating the license plate (7); • the detection on each image (11) of the registration plate (7) of a string of alphanumeric characters (13) of the registration plate (7); • correction of a parallax by the processing unit (6) of the device (2) on each of the images (11) of license plates (7) transmitted; • the reconstruction of the alphanumeric character string (13) identified on each image (11) by optical character recognition; • storage in memory by the processing unit (6) of the device (2) of the reconstructed alphanumeric character strings (13); the method being characterized in that the parallax correction applied to each of the images (11) of license plates (7) is a function of an angle of incidence (0), the angle of incidence (0) being for each image (11) of license plate (7) between the first angle and the second angle relative to the reference axis (A) of the camera (4), the angle of incidence (0) varying for the correction of a parallax of each character (13) detected on each image (11) of license plate (7), the parallax correction applied comprising an application of the Hough transform followed by a rotation of the image (11) of each plate (7).

2. Method for authenticating plates (7) according to claim 1, characterized in that it comprises a step of searching for each alphanumeric character string (13) reconstituted in a database of registration plates (7) acquired subsequently or previously.

3. Method for authenticating plates (7) according to claim 2, characterized in that the database consists of alphanumeric character strings (13) each associated with an image (11) of a license plate (7) acquired by a camera (4) of the device (2) at an entrance (15) of the parking lot (1) and / or images (11) of license plates (7) acquired by a camera (4) of the device (2) at an exit (17) of the parking lot (1).

4. Method for authenticating plates (7) according to claim 3, characterized in the case where a string of alphanumeric characters (13) stored in memory corresponds to an image (11) of license plates (7) from the database acquired by a camera (4) of the device (2) at the exit (17) of the parking lot (1), said string of characters (13) is erased from the memory storage.

5. Authentication device (2) for license plates (7) of a parking lot (1) configured to implement the method according to any one of claims 1 to 4, comprising: • at least one camera (4) arranged on a traffic lane of the parking lot (1), opposite a plurality of parking spaces (5), so that license plates (7) of at least two vehicles (3) parked in adjacent parking spaces (5) are visible by the camera (4); • a processing unit (6) connected to the camera (4) and comprising a storage memory, the processing unit (6) being configured to process images (11) acquired by the camera (4); the authentication device (2) being configured to correct a parallax induced by an angle of incidence (0) between a license plate (7) and a reference axis (A) of the camera (4).

6. Authentication device (2) for license plates (7) according to claim 5, characterized in that it comprises at least two cameras (4) arranged on a traffic lane of the parking lot (1), each camera facing a plurality of parking spaces (5) on a given side of the traffic lane.

7. Authentication device (2) for license plates (7) according to claim 5 or 6, characterized in that it comprises a camera positioned at an entrance (15) of the parking lot (1) and / or a camera positioned at an exit (17) of the parking lot (1).

8. Authentication device (2) for license plates (7) according to any one of claims 5 to 7, characterized in that the processing unit (6) comprises artificial intelligence based on deep neural networks operating by learning.

Images