METHOD AND SYSTEM FOR CORRECTING A LICENSE PLATE

Description

[0001] The disclosure relates to a method and a system for correcting a vehicle registration number (motor vehicle). background

[0002] Automated text recognition (or optical character recognition, OCR) within images is well known. Text recognition is typically applied to a raster image to identify characters such as letters or numbers. It is used, for example, in the transportation sector to identify the license plate number in photos of vehicles and enable automated evaluation of the license plates, for example, in a toll system or during speed checks. In a toll system, license plate recognition can be used to check whether the vehicle in question is authorized to use the road and whether the corresponding toll fee has been paid correctly. However, text recognition is sometimes prone to errors.Especially with license plates, unfavorable lighting conditions when photographing the vehicle, dirt on the license plate or damage to the license plate can lead to incorrect results in automatic license plate recognition.

[0003] Document WO 2009 / 074436 A1 discloses a method and device for detecting a speeding violation by a vehicle. At a first time, the license plate of a vehicle entering a monitoring section is detected and automatically determined using optical license plate recognition. Using a one-way encryption method, a first coding value associated with the vehicle characteristic is determined. At a second time, the vehicle characteristic of a vehicle leaving the monitoring section is detected and, again, automatically determined using optical license plate recognition. Using the one-way encryption method, a second coding value associated with the vehicle characteristic detected at the second time is determined. The first and second coding values ​​are compared.If the first and second coding values ​​match, an average vehicle speed is determined based on the time required to cover the monitored distance. The average speed is compared with a specified maximum speed, and the vehicle characteristic is only stored if the average speed exceeds the maximum speed.

[0004] Document EP 2 779 138 A1 discloses a method for reading vehicle license plates using OCR in a road network. The method comprises: capturing an image of a vehicle license plate at a first location in the road network, OCR-reading a license plate character string in the image, and storing an OCR dataset including the license plate image, the license plate character string, and at least one confidence measure of the OCR reading in a database; capturing an image of a vehicle license plate at a second location in the road network, OCR-reading a license plate character string in the image, and generating a current OCR dataset including the license plate image, the license plate character string, and at least one confidence measure of the OCR reading;and, if at least one confidence measure of the current OCR data set falls below a first minimum confidence value, selecting at least one stored OCR data set from the database whose license plate image has a similarity exceeding a minimum similarity value and / or the greatest similarity to the license plate image of the current OCR data set, and using the at least one selected OCR data set to improve the license plate character string of the current OCR data set. One or more previous OCR reading results of a vehicle license plate are used to support or improve a current OCR reading result in order to reduce the error rate of the OCR reading process.

[0005] Document EP 2 887 333 A1 discloses a method for providing data protection-compliant evidence when evaluating a vehicle license plate using OCR. At a first location and at a first time, a first image of a vehicle is taken with a camera. In the first image, the vehicle license plate is highlighted, and text recognition of the license plate is performed. At a second location and at a second time, a second image of the vehicle is taken with a different camera. In the second image, text recognition of the vehicle license plate is also performed. Based on the first and second times, an average speed of the vehicle for the distance between the first and second locations is determined and compared with a predetermined maximum speed. If the predetermined maximum speed is exceeded, evidence is secured. For this purpose, a hash value is created from the license plate number and the first time.Known errors can occur during OCR recognition of the license plate number. Possible errors are to be corrected using a specified number of permutations of the characters of the captured license plate number.

[0006] Document EP 2 469 497 A1 describes methods for capturing images of vehicles traveling at excessive speed along a stretch of road between an entrance and an exit.The method comprises: generating a random access identifier for a vehicle; taking an image of the vehicle at the entrance and storing the entry image and the access identifier in a first memory; detecting the entry time and at least one identifier of the vehicle at the entrance; forming an encrypted or hashed value of this entry identifier and storing the entry time, the value, and the access identifier as a data record in a second memory; detecting the exit time and at least one identifier of a vehicle at the exit; forming an encrypted or hashed value of this exit identifier and determining the data record with the same value from the second memory and, if the exit time is within a predetermined time period of the entry time of this data record, using the access identifier from this data record to access the first memory to retrieve the entry image stored therewith.

[0007] Further applications of OCR recognition of license plates are described in documents EP 2 320 384 A1, EP 2 360 647 A1 and EP 2 863 338 A2.

[0008] Text recognition and error correction of electronic texts are, of course, also used outside of the transport sector. The document Furrer, L.; Volk, M. "Reducing OCR errors in Gothicscript documents", in: 8th International Conference on Recent Advances in Natural Language Processing (RANLP 2011), pages 97-103 describes a method for correcting errors in documents written in Gothic script. The document M. Reynaert "Text Induced Spelling Correction" (XP058392742) describes a method for correcting texts using hash values. This uses hash value collisions to find identical or at least similar words in a text. The document Mor, M.; Fraenkel, AS "A hash code method for detecting and correcting spelling errors", Communications of the ACM, 1982, 25th vol., no. 12, pp. 935-938 describes a method for correcting spelling errors in texts using hash values.The three aforementioned documents each concern existing (extensive) texts. A key assumption here is that, for a misspelled word, there are usually several correctly spelled examples of that word in the text. Therefore, matches or similarities between the word being checked and other words in the text are sought in order to determine its correctness or, if necessary, a correction. However, a motor vehicle license plate should uniquely identify the respective vehicle. Duplication of license plates is not permitted. The procedures described in the three aforementioned documents are therefore not applicable for correcting a license plate. Summary

[0009] The objective is to provide improved technologies for text recognition of a motor vehicle license plate. In particular, the goal is to enable automatic correction of incorrect text recognition.

[0010] A method according to claim 1 and a system according to claim 5 are disclosed. Further embodiments are subject to dependent claims.

[0011] One or more (e.g. all) steps of the method can be carried out with one or more processors of a data processing device.

[0012] The mark may be represented by one or more characters of a character set, for example by any combination of letters, numbers, spaces and / or special characters.

[0013] The license plate number can be captured optically, for example, using an image recording device such as a digital camera. For example, a photo can be taken of the front and / or rear of a motor vehicle, depicting the vehicle's license plate number.

[0014] The mark can be in the form of an image file, for example as a raster graphic. Text recognition converts the previously (e.g. optically) captured mark into a text-encoded format. The characters of the mark are identified, and each character is assigned a numerical value that corresponds to the character in a common text encoding. Examples of common text encodings are ASCII (American Standard Code for Information Interchange) and Unicode. Other text encodings can also be used. The result of text recognition is the OCR mark. The OCR mark can also comprise one or more characters, for example a set of N characters (N is a natural number). The OCR mark comprises, at each position, the character that was determined to be the most likely character in the mark during text recognition. Automated text recognition within documents is well known.The details of text recognition will therefore not be discussed further here.

[0015] Errors can occur during text recognition. Some errors can also occur systematically. If a (systematic) error is discovered, it is generally possible to correct the error and store the incorrectly recognized license plate number along with the necessary correction(s) in a correction database. This way, if the same license plate number occurs later and its text recognition is again incorrect, the correction database can be accessed to correct the license plate number. However, particularly with personal data, data protection regulations may stipulate that the date (in this case, the license plate number) or an incorrectly recognized OCR date (here: OCR license plate number) may not be stored (not even for correction purposes).

[0016] According to the invention, a hash value is therefore determined based on the OCR identifier, for example by calculation using a hash function. A hash function (also called a hash value function or one-way hash function) is a one-way function that maps an input set of any size (e.g., plaintext of any length, here the OCR identifier) ​​to a value of fixed length. A one-way function is a function that is easy to calculate (with relatively little computational effort), but whose inverse is impossible or very difficult to calculate (with very high computational effort). The hash value does not allow any conclusions to be drawn about the identifier itself. However, applying the hash function to two identical (OCR) identifiers leads to identical hash values. In the correction database, the identifiers are not stored in plain text, but only their hash values.The hash value of the OCR tag can be used to search the correction database to see if there is a matching hash value.

[0017] At least one correction instruction is stored in the correction database for each stored hash value. The at least one correction instruction specifies how the incorrect OCR identifier is to be modified so that it corresponds to the actual identifier. If the comparison of the hash value generated from the OCR identifier with the correction database is positive, the correction instruction (or instructions) assigned to the stored hash value can be applied to the OCR identifier to generate a corrected OCR identifier that corresponds to the actual identifier.

[0018] It may be specified that the correction instruction covers only a specific part of the OCR mark. For example, the correction instruction cannot cover more than M of N characters where M < N (M is a natural number).

[0019] The at least one correction instruction may have one of the following forms: replace the character "X" with the character "Y", replace the character at position n with the character "Y", or replace the character "X" at position n with the character "Y".

[0020] Here, "X" and "Y" are individual characters of a character set, for example, a letter, a digit, a space, and / or a special character; n is a natural number (1, 2, 3, ...) and indicates the position of the character in the OCR tag. If multiple correction instructions are stored for a hash value, the multiple correction instructions can be any combination of the forms listed above.

[0021] The at least one correction instruction is preferably provided in machine-executable form so that the correction of the OCR mark can be carried out automatically (e.g. by means of a processor) and without user intervention.

[0022] The hash value can be generated using a cryptographic hash function. The hash values ​​of the identifiers in the correction database can each be generated using the cryptographic hash function. A cryptographic hash function (or cryptological hash function) is a special form of a hash function that is collision-resistant. It is practically impossible to find two different input values ​​that result in an identical hash value. Suitable hash functions are SHA-1 (SHA - secure hash algorithm), SHA-2, SHA-3, MD4 (MD - message digest), and MD5. Other hash functions can also be used.

[0023] The comparison of the generated hash value with the correction database can, in principle, be performed for each OCR tag generated from a tag. However, this carries the risk of altering a correct OCR tag due to the correction database. This is explained using an example. According to the example, the first tag is "B TC 123." The first tag is incorrectly recognized by the text recognition system as "B TO 123." It is assumed that an entry for the first tag exists in the correction database. The entry includes the hash value of the incorrect OCR tag (i.e., the hash value for "B TO 123") and a correction instruction (e.g., replace "O" with "C" or replace the fourth character with "C" (spaces are counted)).If a second license plate number of another vehicle is actually "B TO 123" and is correctly recognized as "B TO 123" by the text recognition system after capture, the OCR license plate number would result in the same hash value as the first license plate number. Therefore, there is a possibility that the (correct) OCR license plate number derived from the second license plate number could be incorrectly changed to "B TC 123" after comparison with the correction database. To prevent this, a reference test for the OCR license plate number is performed before the hash value is generated, and the subsequent steps (generating the hash value, comparing the hash value with the correction database, and, if necessary, correcting the OCR license plate number) are only performed if the reference test is negative. If the reference test is positive, it is assumed that the OCR license plate number was correctly determined and no correction is necessary.

[0024] The OCR license plate number is intended to be compared with a booking database as a reference test, whereby the booking database contains stored license plates (in plain text). The subsequent steps (creating the hash value, comparing the hash value with the correction database, and, if necessary, correcting the OCR license plate number) are only carried out if the comparison is negative, i.e., no stored license plate number matching the OCR license plate number is found in the booking database. The license plates stored in the booking database are intended to be assigned to vehicles that have already paid a toll for a toll road or have booked a toll road. The booking database therefore contains a list of license plates of motor vehicles that are authorized to use a road.

[0025] The method may further comprise detecting at least one vehicle feature. In the correction database, each stored hash value is assigned information relating to at least one vehicle feature. The comparison with the correction database is then performed taking the at least one vehicle feature into account. This means that, for the comparison with the correction database, the generated hash value and the at least one detected vehicle feature are compared with the hash values ​​stored in the correction database and the respective associated information relating to the at least one vehicle feature.

[0026] The at least one vehicle feature can be selected from the following group: vehicle type, number of axles of a tractor unit, vehicle height, vehicle width, and vehicle length. Furthermore, the vehicle's color, a geometry (e.g., the front view of a truck is essentially rectangular, the front view of a car is essentially trapezoidal), a profile (side view of the vehicle), a position of the license plate on the vehicle, or special features such as protruding mirrors or superstructures (e.g., a crane or car transporter) can be used as a vehicle feature.

[0027] When taking into account at least one vehicle feature, there are various embodiments for forming the hash value and comparing it with the correction database, which are explained below.

[0028] For the hash value calculation, a hash value can be calculated based on the OCR license plate. In addition, it can be provided that a further hash value is calculated based on the vehicle characteristic (or the vehicle characteristics). Alternatively, it can also be provided that a tuple is formed which includes the OCR license plate and the vehicle characteristic (or the vehicle characteristics). In this case, the hash value is formed based on the tuple. The result of the hash value calculation can therefore be: A) the hash value for the OCR license plate and the vehicle characteristic (or the vehicle characteristics) in plain text, B) the hash value for the OCR license plate and the further hash value for the vehicle characteristic (or the vehicle characteristics), or C) the hash value for the tuple comprising the OCR license plate and the vehicle characteristic (or the vehicle characteristics).

[0029] The correction database is structured based on the hash value calculation. In a first embodiment, the correction database contains several stored hash values ​​of license plates, each together with a vehicle feature (or possibly several vehicle features) in plain text (corresponds to result A of the hash value calculation). In a second embodiment, the correction database contains several stored hash values ​​of license plates, each together with the additional hash values ​​for the vehicle feature (or the vehicle features) (corresponds to result B of the hash value calculation). In a third embodiment, the correction database contains several stored hash values ​​derived from tuples comprising the OCR license plate and the vehicle feature (or the vehicle features) (corresponds to result C of the hash value calculation).

[0030] When considering at least one vehicle feature, the correction database can contain license plate information for each stored hash value (e.g., after one of the hash value calculations A, B, or C). The license plate information comprises either at least one correction instruction (as already described) or a correctness confirmation. The correctness confirmation indicates that the OCR license plate corresponding to the hash value and the vehicle feature (or vehicle features) was correctly recognized by the text recognition system.

[0031] The system may comprise two spatially separated components, for example, a detection device and an evaluation device. The detection device may comprise an image recording unit (e.g., a digital camera) for detecting the license plate. Furthermore, the detection device may comprise a processor and a communication unit. The processor of the detection device may be configured to carry out one or more of the method steps described in the present disclosure. The communication unit may be configured to transmit detection data to the evaluation device and / or receive it from the evaluation device. The detection data may, for example, comprise an image recording with the license plate and / or the OCR license plate. The transmission may be wireless, for example, via a mobile network.

[0032] The evaluation device may comprise a communication unit configured to receive acquisition data from the acquisition device and / or to send it to the acquisition device. The evaluation device may further comprise a processor and a memory. The memory may contain the correction database and / or the booking database. The processor of the evaluation device may be configured to perform one or more of the method steps described in the present disclosure, for example: Calculating a hash value for an OCR identifier, comparing the OCR identifier with the booking database, comparing the hash value of the OCR identifier with the correction database and executing one or more correction instructions to correct the OCR identifier.

[0033] The correction database can be structured as follows, for example. If the comparison of the hash value generated from the OCR identifier with the correction database is negative, i.e., if no stored hash value corresponding to the hash value is available in the correction database, the identifier and the OCR identifier can be forwarded for user-guided further processing. The user compares the identifier and the OCR identifier. If the identifier and the OCR identifier match, nothing further is required; a correction is not necessary. If the identifier and the OCR identifier do not match, the user creates a correction instruction (or several correction instructions, if necessary). Applying the correction instruction (or instructions) creates a corrected OCR identifier so that the corrected OCR identifier corresponds to the actual identifier.The hash value of the incorrect OCR mark is stored in the correction database along with the correction instruction(s). If the OCR mark occurs later, it can then be corrected automatically (without user intervention).

[0034] If at least one vehicle characteristic is recorded in addition to the license plate number, if the OCR license plate number matches the license plate number, a confirmation of correctness can also be stored in the correction database along with the hash value of the OCR license plate number. Based on the example already described, this could look like this: A first vehicle has the license plate number "B TC 123." The color, for example, blue, is recorded as the vehicle parameter. The license plate number is incorrectly recognized as "B TO 123" during text recognition. A second vehicle is yellow and has the license plate number "B TO 123." The license plate number of the second vehicle is correctly recognized as "B TO 123" during text recognition.The correction database can now store that the combination of the hash value of "B TO 123" with the color blue includes a correction instruction, and that the combination of the hash value of "B TO 123" with the color yellow includes a confirmation of correctness, i.e., information that the OCR license plate of the second (yellow) vehicle is correct (and should not be corrected).

[0035] Features disclosed in connection with a method can be applied analogously to a corresponding system and vice versa.

[0036] The terms "motor vehicle" and "vehicle" are used synonymously unless the specific description indicates otherwise. Description of implementation examples

[0037] Exemplary embodiments are explained in more detail below with reference to the figures. They show: Fig. 1 is a schematic representation of a data processing device suitable for carrying out some of the methods disclosed here, in particular those described in the Fig. 2 and Fig. 3 shown methods, Fig. 2 a first embodiment of a method, Fig. 3 a second embodiment of the method, Fig. 4 a schematic representation of a system suitable for carrying out some of the methods disclosed here, in particular those in the Fig. 5 - 10 shown method, Fig. 5 - 10 further embodiments of the method, Fig. 11 a schematic representation of another system suitable for carrying out some of the methods disclosed here, in particular those in the Fig. 12 and Fig. 13 shown method, Fig. 12 a further embodiment of the method and Fig. 13 yet another embodiment of the method.

[0038] In the following, the same reference symbols are used for the same components / process steps.

[0039] In some embodiments, exemplary embodiments of a general date are described. Here, the term "date" can refer to information or details about a person or thing. The date can be a person-related date and / or an object-related date. The date can be represented by one or more characters from a character set, for example, by any combination of letters, numbers, spaces, and / or special characters. The date can be a license plate number of a motor vehicle (the license plate number can be used to identify the owner of the vehicle, thus it is a person-related date). The date can also include other information about a person, such as name, address, telephone number, or date of birth. The date can also be a serial number of a product. Other embodiments refer directly to a license plate number of a motor vehicle.

[0040] A schematic representation of a data processing device 1 is shown in Fig. 1 shown.

[0041] The data processing device 1 comprises an image recording device 2, a processor 3, and a memory 4. The data processing device 1 can be implemented, for example, as a smartphone or a tablet.

[0042] The image recording device 2 is configured to capture an image of a date. The image recording device 2 can be implemented, for example, as a digital camera. A photo of the date is then available as an image file, typically as a raster graphic.

[0043] Processor 3 is configured to perform automatic text recognition (OCR) based on the image of the date and generate an OCR date. The OCR date contains text-encoded individual characters of the date (e.g., letters, numbers, special characters, and / or spaces). The processor can be implemented, for example, as a CPU (central processing unit) or as a microcontroller.

[0044] Memory 4 includes a correction database. Hash values ​​of data and correction instructions are stored in the correction database. Each stored hash value is assigned at least one correction instruction. The at least one correction instruction specifies how an incorrect OCR date is to be changed so that it corresponds to the actual date. Memory 4 can be implemented, for example, as a hard disk or as a flash memory.

[0045] Processor 3 is further configured to calculate a hash value from the OCR data using a hash function. Any known hash function can be used to calculate the hash value, for example, SHA-1, SHA-2, SHA-3, MD4, or MD5.

[0046] The processor 3 is further configured to compare the calculated hash value with the hash values ​​stored in the correction database and, if the comparison is positive, to apply the at least one correction instruction associated with the stored hash value to the OCR datum in order to generate a corrected OCR datum.

[0047] The data processing device 1 can optionally have a display device or be coupled to a display device (not shown). The display device can be configured to display the following elements individually or in any combination: the image of the date, the OCR date generated from the date, the at least one correction instruction, and / or the corrected OCR date.

[0048] In Fig. 2 an embodiment of a method for correcting a date is shown, which does not fall within the scope of the claims.

[0049] First, a date is optically captured using image capture device 2, and an image of the date is created (step 10), for example, in the form of a raster graphic. Processor 3 executes a text recognition program to generate an OCR date from the image of the date (step 20). A hash value is calculated from the OCR date using processor 3 (step 40). The calculated hash value is compared with a correction database using processor 3 (step 50). If a stored hash value corresponding to the calculated hash value is found in the correction database, the corresponding correction instruction(s) are applied to the OCR date using processor 3, and a corrected OCR date is generated (step 60). The corrected OCR date then corresponds to the actual date shown in the image.If the comparison of the calculated hash value with the correction database is negative, the unchanged OCR date is forwarded for user-assisted further processing (step 51). The user-assisted further processing may include a user-assisted comparison of the date with the OCR date and, if necessary, a user-assisted correction of the OCR date, as well as the creation of a new entry in the correction database for the OCR date with at least one correction instruction.

[0050] In Fig. 3 a further embodiment of the method is shown, which does not fall within the scope of the claims.

[0051] According to this embodiment, after capturing the date and creating the image of the date (step 10), a confidence value is determined using processor 3 (step 90) during the OCR recognition of the date (step 20). The confidence value is compared with a predetermined threshold value by processor 3 (step 120).

[0052] If the confidence value is lower than the threshold, this indicates that the OCR recognition was not reliable. In this case, analogous to the procedure in Fig. 1 illustrated embodiment - a hash value is calculated based on the OCR date (step 40), the calculated hash value is compared with the correction database (step 50) and a correction of the OCR date is carried out (step 60) or a further evaluation is initiated (step 51).

[0053] If the confidence value is greater than or equal to the threshold, this indicates that the OCR recognition is correct and that the OCR date matches the date shown in the figure. In this case, the OCR date is used directly (i.e., without comparison with the correction database) for further processing (step 121). Further processing may include machine-assisted (automatic) processing in a data processing system.

[0054] The confidence value is preferably evaluated before calculating the hash value. This can save resources if the confidence value determines that the OCR recognition was likely reliable, so that correction is not necessary and thus the hash value can be omitted.

[0055] Fig. 4shows a schematic representation of a system configured to carry out some of the embodiments of the method disclosed herein.

[0056] The system comprises a detection device 100 and an evaluation device 110. The evaluation device 110 is spatially separated from the detection device 100.

[0057] The detection device 100 has a camera 101 (e.g., a digital camera) for capturing an image of a license plate. The image of the license plate may be part of an image of a motor vehicle, for example, a truck. The image may be generated by photographing the front of the motor vehicle (where the license plate is typically located).

[0058] Furthermore, the detection device 100 comprises a processor 102 (e.g., a CPU or a microcontroller) and a communication unit 103. The detection device 100 can optionally comprise a further communication unit 104. The detection device 100 can be designed as a control bridge spanning the road or as a control column positioned at the side of the road.

[0059] The evaluation device 110 comprises a communication unit 112, a processor 111 (e.g., a CPU or a microcontroller), and a memory 113 (e.g., a hard disk or a flash memory). The evaluation device 110 can be implemented as a server.

[0060] The communication unit 103 of the detection device 100 and the communication unit 112 of the evaluation device 110 are configured to wirelessly exchange data and / or signals with each other in both directions, for example, via a radio interface such as mobile radio, e.g., GSM, UMTS, or LTE (GSM - Global System for Mobile Communications, UMTS - Universal Mobile Telecommunications System, LTE - Long Term Evolution). Data and / or signals can be sent from the communication unit 103 of the detection device 100 to the communication unit 112 of the evaluation device 110. It is also possible for data and / or signals to be sent from the communication unit 112 of the evaluation device 110 to the communication unit 103 of the detection device 100.

[0061] The further communication unit 104 of the detection device 100 is configured to exchange data and / or signals via a radio interface with a vehicle device arranged in a vehicle, i.e., to send data and / or signals to the vehicle device and / or receive them from the vehicle device. In particular, the further communication unit 104 of the detection device 100 can be configured to send a request to the vehicle device, requesting the vehicle device to transmit vehicle data to the further communication unit 104 of the detection device 100. The vehicle data can include the vehicle license plate number. The radio interface of the further communication unit 104 of the detection device 100 can be implemented as a DSRC interface (DSRC - dedicated short range communication). DSRC enables wireless communication in a very small communication zone (range typically up to 50 m).DSRC can be deployed as infrared DSRC or microwave DSRC (µWave DSRC). Microwave DSRC, for example, transmits at a frequency in the 5.8–5.9 GHz range with a maximum transmission power of 2 W. In Europe, microwave DSRC is used, among other things, for electronic toll collection (Austria, Poland, the Czech Republic, and France) and for monitoring GNSS-based toll collection devices (Germany, Slovakia, and Belgium).

[0062] The memory 113 of the evaluation device 110 can contain a correction database and / or a booking database. The correction database and the booking database can also be stored on two different storage media. It can also be provided that the correction database and the booking database are implemented in different computer systems (each with its own processors and memory elements) (not shown).

[0063] A further embodiment of the method is described in Fig. 5 shown.

[0064] The camera 101 of the detection device 100 takes a photo which includes a license plate number of a motor vehicle (step 10).

[0065] The photo is transmitted to the communication unit 112 of the evaluation device 110 using the communication unit 103 of the capture device 100 (step 11).

[0066] Using processor 111 of evaluation device 110, an OCR code is created based on the code shown in the photo (step 20). The OCR code is then available in text-encoded form and can be further processed by processor 111 of evaluation device 110.

[0067] The OCR license plate is compared with a booking database using processor 111 of evaluation device 110 (step 30). The booking database contains stored vehicle license plates (in plain text). The stored license plates are assigned to toll bookings or toll exemptions, i.e., they represent vehicles for which either a toll-based route has been booked and paid, or vehicles that are exempt from toll payment.

[0068] If an entry is found in the booking database for the OCR license plate (step 31), there are two possibilities: i) the OCR license plate was correctly determined, and a proper booking or toll exemption exists, or ii) the OCR license plate was incorrectly determined (e.g., the actual license plate number "B TC 123" was recognized as "B TO 123"); however, there is a booking or toll exemption for another vehicle (which actually has the license plate number "B TO 123") for the incorrectly recognized OCR license plate (false positive result). The method can optionally include a further step in which information is sent from the evaluation device 110 to the recording device 100 that an entry has been found in the booking database for the license plate (not shown).

[0069] When the method is applied in a toll system, the check is performed position-based. The position of the detection device 100 is assigned to a section of road. The time of detection of the license plate number is also taken into account. For a false positive result to occur, two vehicles (with the license plates "B TC 123" and "B TO 123") would have to be at almost the same location (the position of the detection device 100) at almost the same time (within a few hours). This scenario is considered very unlikely and will not be pursued further here.

[0070] If the comparison of the OCR license plate with the booking database does not yield a result, it is possible that the OCR recognition of the license plate did not produce the correct license plate. A hash value is then calculated from the OCR license plate using the processor 111 of the evaluation device 110 (step 40).

[0071] The calculated hash value is compared with a correction database using processor 111 of evaluation device 110 (step 50). The correction database contains several stored hash values ​​of license plates and at least one correction instruction for each stored hash value. The correction instruction specifies how the OCR license plate must be modified so that it corresponds to the actual license plate. Based on the above example, the correction instruction can be, for example: i) replace the character "O" with the character "C" or ii) replace the fourth character (position n = 4, spaces are included) of the license plate with the character "C".

[0072] If the comparison with the correction database does not produce a match, the OCR mark is forwarded for user-assisted further processing (step 51).

[0073] If an entry is found in the correction database for the hash value, the correction instruction (or correction instructions) associated with the stored hash value are applied to the OCR identifier by means of the processor 111 of the evaluation device 110 and a corrected OCR identifier is generated (step 60).

[0074] The corrected OCR license plate is then compared with the booking database by processor 111 of evaluation device 110 (step 70). If there is no match, user-assisted further processing takes place (step 71). If a match is found, it is assumed that a correct booking or toll exemption has been made (step 80). The method can optionally include a further step in which information is sent from evaluation device 110 to recording device 100 indicating that an entry for the license plate has been found in the booking database (not shown).

[0075] A further embodiment of the method is described in Fig. 6 which does not fall within the scope of the claims.

[0076] The camera 101 of the detection device 100 takes a photo which includes a license plate number of a motor vehicle (step 10).

[0077] The photo is transmitted to the communication unit 112 of the evaluation device 110 using the communication unit 103 of the capture device 100 (step 11).

[0078] Using processor 111 of evaluation device 110, an OCR code is created based on the code shown in the photo (step 20). The OCR code is then available in text-encoded form and can be further processed by processor 111 of evaluation device 110.

[0079] During text recognition of the license plate, one or more confidence values ​​are calculated by processor 111 of evaluation device 110 (step 90). A confidence value is a measure of the probability that the result of the text recognition corresponds to the original source text. The confidence value can be determined for a word (a complete character string such as a complete license plate) or for individual characters. Preferably, in the embodiments disclosed here, a confidence value is calculated for each individual character (a letter, a number, a space, and / or a special character) of the license plate. Therefore, for the exemplary license plate "B TC 123," eight confidence values ​​(for three letters, three numbers, and two spaces) are determined.

[0080] The confidence value or confidence values ​​are compared with a predetermined threshold value by means of the processor 111 of the evaluation device 110 (step 120).

[0081] If the confidence value or one of the multiple confidence values ​​is lower than the threshold, this indicates that text recognition was unsuccessful for at least part of the license plate, and thus the OCR license plate does not correspond to the actual license plate. In this case, a hash value is calculated from the OCR license plate using the processor 111 of the evaluation device 110 (step 40).

[0082] The calculated hash value is compared with a correction database using processor 111 of evaluation device 110 (step 50). If the comparison with the correction database does not result in a match, the OCR identifier is forwarded for user-assisted further processing (step 51). If an entry is found for the hash value in the correction database, the correction instruction (or instructions) associated with the stored hash value are applied to the OCR identifier using processor 111 of evaluation device 110, and a corrected OCR identifier is generated (step 60).

[0083] The corrected OCR license plate is then compared with the booking database by processor 111 of evaluation device 110 (step 70). If there is no match, user-assisted further processing takes place (step 71). If a match is found, it is assumed that a correct booking or toll exemption has been made (step 80). The method can optionally include a further step in which information is sent from evaluation device 110 to recording device 100 indicating that an entry for the license plate has been found in the booking database (not shown).

[0084] If the confidence value or all of the multiple confidence values ​​are equal to or greater than the threshold value, this indicates that the text recognition of the license plate was successful and thus the OCR license plate corresponds to the actual license plate. The OCR license plate is compared with the booking database using the processor 111 of the evaluation device 110 (step 30). If an entry is found in the booking database for the OCR license plate (step 31), there are - as already mentioned - two possibilities: i) the OCR license plate was correctly determined and a proper booking or toll exemption has been made, or ii) a false positive result. The method can optionally include a further step in which information is sent from the evaluation device 110 to the recording device 100 that an entry was found in the booking database for the license plate (not shown).

[0085] If the comparison of the OCR license plate with the booking database yields no result, despite the high confidence value (or high confidence values), there is a possibility that the OCR recognition of the license plate did not produce the correct license plate. A hash value is then calculated from the OCR license plate using the processor 111 of the evaluation device 110 (step 40).

[0086] The calculated hash value is compared with a correction database using processor 111 of evaluation device 110 (step 50). If the comparison with the correction database does not result in a match, the OCR identifier is forwarded for user-assisted further processing (step 51). If an entry is found in the correction database for the calculated hash value, the correction instruction (or instructions) associated with the stored hash value are applied to the OCR identifier using processor 111 of evaluation device 110, and a corrected OCR identifier is generated (step 60).

[0087] The corrected OCR license plate is then compared with the booking database by processor 111 of evaluation device 110 (step 70). If there is still no match, user-assisted further processing occurs (step 71). If a match is found, it is assumed that a correct booking or toll exemption exists (step 80). The method can optionally include a further step in which information is sent from evaluation device 110 to recording device 100 indicating that an entry for the license plate has been found in the booking database (not shown).

[0088] Fig. 7 shows another embodiment of the method.

[0089] A photograph is taken using camera 101 of capture device 100, which includes a license plate of a motor vehicle (step 10). Using processor 102 of capture device 100, an OCR license plate is created based on the license plate depicted in the photograph (step 20). The OCR license plate is transmitted to communication unit 112 of evaluation device 110 using communication unit 103 of capture device 100 (step 12).

[0090] The OCR license plate is compared with a booking database using processor 111 of evaluation device 110 (step 30). If an entry is found in the booking database for the OCR license plate (step 31), there are two possibilities, as already mentioned: i) the OCR license plate was correctly determined and a proper booking or toll exemption has been made, or ii) a false positive result. The method can optionally include a further step in which information is sent from evaluation device 110 to recording device 100 indicating that an entry for the license plate was found in the booking database (not shown).

[0091] If the comparison of the OCR license plate with the booking database does not yield a result, it is possible that the OCR recognition of the license plate did not produce the correct license plate. A hash value is calculated from the OCR license plate using the processor 111 of the evaluation device 110 (step 40).

[0092] The calculated hash value is compared with a correction database using processor 111 of evaluation device 110 (step 50). If the comparison with the correction database does not result in a match, the OCR mark is forwarded for user-assisted further processing (step 51). User-assisted further processing may include requesting the photo with the mark in order to perform a comparison between the mark and the OCR mark. If an entry is found in the correction database for the hash value, the correction instruction (or instructions) associated with the stored hash value are applied to the OCR mark by processor 111 of evaluation device 110 to generate a corrected OCR mark (step 60).

[0093] The corrected OCR license plate is then compared with the booking database by processor 111 of evaluation device 110 (step 70). If there is no match, user-assisted further processing takes place (step 71). If a match is found, it is assumed that a correct booking or toll exemption has been made (step 80).

[0094] The method may optionally comprise a further step in which information is sent from the evaluation device 110 to the recording device 100 that an entry for the license plate has been found in the booking database (not shown).

[0095] In an alternative embodiment, in step 12, the OCR identifier is transmitted together with the photo from the capture device 100 to the evaluation device 110. In this case, requesting the photo is not necessary during user-assisted further processing (step 51).

[0096] In Fig. 8 a further embodiment of the method is shown, which does not fall within the scope of the claims.

[0097] The camera 101 of the capture device 100 takes a photograph containing a motor vehicle license plate (step 10). The processor 102 of the capture device 100 creates an OCR license plate based on the license plate imaged in the photograph (step 20).

[0098] During the text recognition of the license plate, one or more confidence values ​​are calculated by means of the processor 102 of the detection device 100 (step 90).

[0099] The OCR identifier and the confidence value (or confidence values) are transmitted to the communication unit 112 of the evaluation device 110 using the communication unit 103 of the capture device 100 (step 14).

[0100] The confidence value or confidence values ​​are compared with a predetermined threshold value by means of the processor 111 of the evaluation device 110 (step 120).

[0101] If the confidence value or one of the multiple confidence values ​​is lower than the threshold, this indicates that the text recognition of the license plate was unsuccessful and thus the OCR license plate does not correspond to the actual license plate. In this case, a hash value is calculated based on the OCR license plate using the processor 111 of the evaluation device 110 (step 40).

[0102] The calculated hash value is compared with a correction database using processor 111 of evaluation device 110 (step 50). If the comparison with the correction database does not result in a match, the OCR mark is forwarded for user-assisted further processing (step 51). User-assisted further processing may include requesting the photo with the mark in order to perform a comparison between the mark and the OCR mark. If an entry is found in the correction database for the calculated hash value, the correction instruction (or instructions) associated with the stored hash value are applied to the OCR mark by processor 111 of evaluation device 110, and a corrected OCR mark is generated (step 60).

[0103] The corrected OCR license plate is then compared with the booking database by processor 111 of evaluation device 110 (step 70). If there is no match, user-assisted further processing takes place (step 71). If a match is found, it is assumed that a correct booking or toll exemption has been made (step 80). The method can optionally include a further step in which information is sent from evaluation device 110 to recording device 100 indicating that an entry for the license plate has been found in the booking database (not shown).

[0104] If the confidence value or all of the multiple confidence values ​​are equal to or greater than the threshold value, this is an indication that the text recognition of the license plate number was successful and thus the OCR license plate number corresponds to the actual license plate number. The OCR license plate number is compared with the booking database using the processor 111 of the evaluation device 110 (step 30). If an entry is found in the booking database for the OCR license plate number (step 31), there are two possibilities: i) the OCR license plate number was correctly determined and a proper booking or toll exemption has been made, or ii) a false positive result. The method can optionally include a further step in which information is sent from the evaluation device 110 to the recording device 100 that an entry was found in the booking database for the license plate number (not shown).

[0105] If the comparison of the OCR license plate with the booking database yields no result, despite the high confidence value (or high confidence values), there is a possibility that the OCR recognition of the license plate did not produce the correct license plate. A hash value is calculated from the OCR license plate using the processor 111 of the evaluation device 110 (step 40).

[0106] The calculated hash value is compared with a correction database using processor 111 of evaluation device 110 (step 50). If the comparison with the correction database does not result in a match, the OCR identifier is forwarded for user-assisted further processing (step 51). If an entry is found in the correction database for the calculated hash value, the correction instruction (or instructions) associated with the stored hash value are applied to the OCR identifier using processor 111 of evaluation device 110, and a corrected OCR identifier is generated (step 60).

[0107] The corrected OCR license plate is then compared with the booking database by processor 111 of evaluation device 110 (step 70). If there is still no match, user-assisted further processing takes place (step 71). If a match is found, it is assumed that a correct booking or toll exemption has been made (step 80). The method can optionally include a further step in which information is sent from evaluation device 110 to recording device 100 indicating that an entry for the license plate has been found in the booking database (not shown).

[0108] In an alternative embodiment, in step 14, the OCR identifier and the confidence value(s) are transmitted together with the photo from the capture device 100 to the evaluation device 110. In this case, requesting the photo is not necessary during user-assisted further processing (step 51).

[0109] A further embodiment of the method is described in Fig. 9 which does not fall within the scope of the claims.

[0110] The camera 101 of the capture device 100 takes a photograph containing a motor vehicle license plate (step 10). The processor 102 of the capture device 100 creates an OCR license plate based on the license plate imaged in the photograph (step 20).

[0111] The further communication unit 104 of the detection device 100 receives vehicle data from an on-board device arranged in the motor vehicle (step 130). The vehicle data includes a reference license plate. The reference license plate is manually entered by a user of the motor vehicle when the on-board device is set up. When used as intended (i.e., the vehicle license plate is entered correctly), the reference license plate corresponds to the actual license plate of the motor vehicle. Receipt of the vehicle data can be preceded by a request from the further communication unit 104 to the on-board device, which initiates the transmission of the vehicle data.

[0112] The order of steps 10, 20, and 130 is not important. Only step 10 must necessarily occur before step 20. The three steps are usually performed within a relatively short period of time (on the order of a few milliseconds to approximately 2 seconds). The order can depend on the road conditions, weather conditions, and / or the speed of the motor vehicle. However, step 130 can occur before step 10, between step 10 and step 20, after step 20, or simultaneously with one of steps 10 or 20. After the three steps have been performed, the OCR license plate and the reference license plate are available in the capture device 100 and can be processed there.

[0113] The OCR license plate is compared with the reference license plate by processor 102 of the detection device 100 (step 140). If the comparison is positive, i.e., the OCR license plate and the reference license plate match, the method is completed (step 141). It is assumed that the vehicle's on-board unit is set up correctly and the driver is participating in the toll collection process. It is possible that the reference license plate was mistakenly or intentionally entered incorrectly during the setup of the on-board unit and thus does not correspond to the actual license plate, and that the OCR recognition of the license plate results in an OCR license plate that matches the incorrect reference license plate. In this case, correct participation in the toll collection process would be erroneously assumed. However, this scenario is considered extremely unlikely and will not be considered further here.

[0114] If the OCR identifier and the reference identifier do not match, a hash value is calculated by the processor 102 of the capture device 100 based on the OCR identifier (step 40).

[0115] The calculated hash value is transmitted to the communication unit 112 of the evaluation device 110 using the communication unit 103 of the detection device 100 (step 150).

[0116] The calculated hash value is compared with a correction database using processor 111 of evaluation device 110 (step 50). If the comparison with the correction database does not result in a match, the OCR license plate is forwarded for user-assisted further processing (step 51). User-assisted further processing may include sending the photo of the license plate, the OCR license plate, and / or the reference license plate from capture device 100 to evaluation device 110.

[0117] If an entry is found in the correction database for the calculated hash value, the correction instruction (or correction instructions) associated with the stored hash value are transmitted via the communication unit 112 of the evaluation device 110 to the communication unit 103 of the detection device 100 (step 160).

[0118] The correction instruction (or correction instructions) is applied to the OCR mark by means of the processor 102 of the capture device 100 and a corrected OCR mark is generated (step 60).

[0119] The processor 102 of the detection device 100 compares the corrected OCR license plate with the reference license plate (step 170). If the comparison is positive, i.e., the corrected OCR license plate and the reference license plate match, the method is completed (step 171). It is assumed that the vehicle's on-board device is correctly configured and the driver is participating in the toll collection process.

[0120] If the corrected OCR mark and the reference mark do not match, the process proceeds to user-assisted further processing (step 180). User-assisted further processing may include sending the photo of the mark, the OCR mark, the corrected OCR mark, and / or the reference mark from the capture device 100 to the evaluation device 110.

[0121] After step 20, the method may optionally include calculating one or more confidence values ​​for the text recognition of the license plate using the processor 102 of the detection device 100. In the case of user-assisted further processing, the confidence value (or confidence values) may be an indication of the reliability of the text recognition.

[0122] Fig. 10 shows a further embodiment of the method which does not fall within the scope of the claims.

[0123] The camera 101 of the detection device 100 takes a photo which includes a license plate number of a motor vehicle (step 10).

[0124] Using the processor 102 of the capture device 100, an OCR license plate is created based on the photo of the license plate (step 20).

[0125] The further communication unit 104 of the detection device 100 receives vehicle data from an on-board device arranged in the motor vehicle (step 130). The vehicle data includes a reference license plate. The reference license plate is manually entered by a user when setting up the on-board device. When used as intended, the reference license plate corresponds to the actual license plate of the motor vehicle. Receipt of the vehicle data can be preceded by a request from the further communication unit 104 to the on-board device, which initiates the transmission of the vehicle data.

[0126] As in the embodiment according to Fig. 9 The order of steps 10, 20, and 130 is irrelevant. After the three steps have been performed in any order, the OCR identifier and the reference identifier are available in the capture device 100 and can be processed there.

[0127] The OCR identifier is compared with the reference identifier by processor 102 of the capture device 100 (step 140). If the comparison is positive, i.e., the OCR identifier and the reference identifier match, the method is completed (step 141).

[0128] If the OCR identifier and the reference identifier do not match, the OCR identifier and the reference identifier are transmitted to the communication unit 112 of the evaluation device 110 using the communication unit 103 of the capture device 100 (step 155).

[0129] Using the processor 111 of the evaluation device 110, a hash value is calculated based on the OCR identifier (step 40).

[0130] The calculated hash value is compared with a correction database using the processor 111 of the evaluation device 110 (step 50).

[0131] If the comparison with the correction database does not produce a match, the OCR license plate is forwarded for user-assisted further processing (step 51). User-assisted further processing may include sending the photo with the license plate from the capture device 100 to the evaluation device 110.

[0132] If an entry is found in the correction database for the calculated hash value, the correction instruction (or correction instructions) associated with the stored hash value are applied to the OCR identifier by means of the processor 111 of the evaluation device 110 and a corrected OCR identifier is generated (step 60).

[0133] The corrected OCR identifier is compared with the reference identifier by means of the processor 111 of the evaluation device 110 (step 175).

[0134] If the comparison is positive, i.e., the corrected OCR license plate and the reference license plate match, the method is completed (step 171). It is assumed that the vehicle's on-board device is correctly configured and the vehicle is participating in the toll collection process. The method can optionally include a step in which the evaluation device 110 transmits information about the positive result of the comparison to the detection device 100.

[0135] If the corrected OCR identifier and the reference identifier do not match, the process proceeds to user-assisted further processing (step 180). User-assisted further processing may include sending the photo with the identifier from the capture device 100 to the evaluation device 110. The method may optionally include a step in which the evaluation device 110 transmits information about the negative result of the comparison to the capture device 100.

[0136] After step 20, the method may optionally include calculating one or more confidence values ​​for the text recognition of the license plate using the processor 102 of the detection device 100. In the case of user-assisted further processing, the confidence value (or confidence values) may be an indication of the reliability of the text recognition.

[0137] In an alternative embodiment, in step 155, the OCR identifier and the reference identifier are transmitted from the capture device 100 to the evaluation device 110 together with the photo of the identifier. In this case, transmitting the photo is not required during user-assisted further processing (steps 51 or 171).

[0138] The embodiments disclosed here carry the risk (which is small in practice) of altering a correct license plate number due to the correction database. According to the above example, a license plate number "B TC 123" may be incorrectly recognized by the OCR as "B TO 123." If a user-assisted correction is performed and a correction instruction is created, the correction database will then contain the hash value for the incorrect OCR license plate number ("B TO 123") along with a correction instruction (e.g., replace "O" with "C" or replace the fourth character with "C"). If a license plate number of another vehicle is actually "B TO 123," is captured, and subsequently correctly recognized by the OCR as "B TO 123," there is a possibility that it will be changed to "B TC 123" after comparison with the correction database.

[0139] In order to avoid such errors, a further embodiment provides that at least one vehicle feature is recorded together with the license plate and taken into account when comparing with the correction database.

[0140] Fig. 11 shows a schematic representation of another embodiment of a system configured to carry out some of the embodiments of the method disclosed herein.

[0141] The system comprises a detection device 200 and an evaluation device 110. The evaluation device 110 is spatially separated from the detection device 200.

[0142] The detection device 200 has a camera 201 (e.g., a digital camera) for capturing an image of a license plate. The image of the license plate may be part of an image of a motor vehicle, for example, a truck. The image may be generated by photographing the front of the motor vehicle (where the license plate is typically located).

[0143] Furthermore, the detection device 200 has a processor 202 (e.g., a CPU or a microcontroller), a communication unit 203, and a detection unit 205.

[0144] The detection unit 205 is configured to determine a vehicle feature. The vehicle feature can be the vehicle type, the number of axles of a tractor, a vehicle dimension such as the length, width, or height of the vehicle, the color of the vehicle, the position of the license plate on the vehicle, and any combination of the aforementioned features. The vehicle type can be determined, for example, based on the geometry of the vehicle, e.g., from the front view of the vehicle and / or from the profile of the vehicle. Special features of a vehicle, such as protruding mirrors or superstructures such as those of a crane or a car transporter, can also be detected as a vehicle feature using the detection unit 205. The detection unit 205 can be configured as a camera (e.g., a digital camera), a laser scanner (LIDAR), a time-of-flight camera, a stereo camera, or a combination thereof.

[0145] The detection device 200 may optionally have a further communication unit 204.

[0146] The detection device 200 can be designed as a control bridge spanning the road or as a control column positioned at the side of the road.

[0147] The evaluation device 110 comprises a communication unit 112, a processor 111 (e.g., a CPU or a microcontroller), and a memory 113 (e.g., a hard disk or a flash memory). The evaluation device 110 can be implemented as a server.

[0148] The communication unit 203 of the detection device 200 and the communication unit 112 of the evaluation device 110 can be configured to wirelessly exchange data and / or signals with each other in both directions, for example, via a radio interface such as mobile radio, e.g., GSM, UMTS, or LTE (GSM - Global System for Mobile Communications, UMTS - Universal Mobile Telecommunications System, LTE - Long Term Evolution). Data and / or signals can be sent from the communication unit 203 of the detection device 200 to the communication unit 112 of the evaluation device 110. It is also possible for data and / or signals to be sent from the communication unit 112 of the evaluation device 110 to the communication unit 203 of the detection device 200.

[0149] The further communication unit 204 of the detection device 200 can be configured to exchange data and / or signals via a radio interface with a vehicle device arranged in a vehicle, i.e., to send data and / or signals to the vehicle device and / or receive them from the vehicle device. In particular, the further communication unit 204 of the detection device 200 can be configured to send a request to the vehicle device, requesting the vehicle device to transmit vehicle data to the further communication unit 204 of the detection device 200. The vehicle data can include the vehicle license plate number. The radio interface of the further communication unit 204 of the detection device 200 can be implemented as a DSRC interface.

[0150] The memory 113 of the evaluation device 110 can contain a correction database and / or a booking database. The correction database and the booking database can also be stored on two different storage media. It can also be provided that the correction database and the booking database are implemented in different computer systems (each with its own processors and memory elements) (not shown).

[0151] Vehicle features that can be determined from the front view of the vehicle, in particular the vehicle type, a vehicle dimension such as the width or height of the vehicle, the color of the vehicle, the position of the license plate on the vehicle, protruding mirrors and any combination of the aforementioned features, can also be determined by means of the camera 201. To determine these vehicle features, the detection unit 205 is not required. For these vehicle features, the following with reference to the Fig. 12 and 13 described procedures also with the system according to Fig. 4 be executed.

[0152] Fig. 12 shows another embodiment of the method.

[0153] The camera 201 of the detection device 200 takes a photo which includes a license plate number of a motor vehicle (step 10).

[0154] A vehicle feature of the motor vehicle is detected by the detection unit 205 of the detection device 200 (step 210). In step 210, multiple vehicle features can also be detected. For example, a laser scan of the vehicle can be performed while the vehicle passes the detection device 200 (e.g., passes under or past the detection device 200).

[0155] The order of steps 10 and 210 is not important. Step 210 can also be executed before step 10. It is also possible for both steps 10 and 210 to be executed simultaneously.

[0156] The photo and the vehicle feature (or vehicle features) are transmitted to the communication unit 112 of the evaluation device 110 using the communication unit 203 of the detection device 200 (step 220).

[0157] Using processor 111 of evaluation device 110, an OCR code is created based on the photo of the license plate (step 20). The OCR code is then available in text-encoded form and can be further processed by processor 111 of evaluation device 110.

[0158] The OCR license plate is compared with a booking database using processor 111 of evaluation device 110 (step 30). The booking database contains stored vehicle license plates (in plain text). If an entry is found in the booking database for the OCR license plate (step 31), there are two possibilities: i) the OCR license plate was correctly determined and a proper booking or toll exemption exists, or ii) a false positive result. The method can optionally include a further step in which information is sent from evaluation device 110 to recording device 200 that an entry for the license plate was found in the booking database (not shown).

[0159] If the comparison of the OCR license plate with the booking database does not produce a result, it is possible that the OCR recognition of the license plate did not produce the correct license plate.

[0160] A hash value calculation is carried out by means of the processor 111 of the evaluation device 110 (step 230). For the hash value calculation, a hash value is calculated based on the OCR license plate. Additionally, it can be provided that a further hash value is calculated based on the vehicle feature (or the vehicle features). Alternatively, it can also be provided that a tuple is formed which comprises the OCR license plate and the vehicle feature (or the vehicle features). In this case, the hash value is formed based on the tuple. The result of the hash value calculation is therefore: A) the hash value for the OCR license plate and the vehicle feature (or the vehicle features) in plain text, B) the hash value for the OCR license plate and the further hash value for the vehicle feature (or the vehicle features), or C) the hash value for the tuple comprising the OCR license plate and the vehicle feature (or the vehicle features).

[0161] In the next step, the result of the hash value calculation is compared with a correction database using processor 111 of evaluation device 110 (step 240). The correction database is structured similarly to the hash value calculation. In a first embodiment, the correction database contains several stored hash values ​​of license plates, each together with a vehicle characteristic (or possibly several vehicle characteristics) in plain text (corresponds to result A of the hash value calculation). Each stored hash value with a vehicle characteristic is assigned a piece of license plate information. The license plate information comprises either at least one correction instruction (as already described in the present application) or a correctness confirmation. The correctness confirmation indicates that the OCR license plate corresponding to the calculated hash value and the vehicle characteristic (or vehicle characteristics) was correctly recognized by the text recognition.In a second embodiment, the correction database contains a plurality of stored hash values ​​of license plates, each together with the further hash value for the vehicle feature (or features) (corresponds to result B of the hash value calculation). Each pair consisting of the stored hash value and the further hash value is in turn assigned a piece of license plate information. In a third embodiment, the correction database contains a plurality of stored hash values ​​derived from tuples comprising the OCR license plate and the vehicle feature (or features) (corresponds to result C of the hash value calculation). Each stored hash value of the tuples is assigned a piece of license plate information. The statements regarding the license plate information apply accordingly to the second and third embodiments.

[0162] If the comparison with the correction database fails to match, or if the comparison is positive and the OCR license plate is confirmed as correct, the OCR license plate is forwarded for user-assisted processing (step 51). If the OCR license plate is correct and the license plate is not found in the booking database, this may be an attempt at toll fraud, which will be investigated.

[0163] If the comparison is positive and a correction instruction (or several correction instructions) for the OCR identifier are present, the correction instruction (or the correction instructions) are applied to the OCR identifier using the processor 111 of the evaluation device 100 and a corrected OCR identifier is generated (step 60).

[0164] The corrected OCR license plate is then compared with the booking database by processor 111 of evaluation device 110 (step 70). If there is no match, user-assisted further processing takes place (step 71). If a match is found, it is assumed that a correct booking or toll exemption has been made (step 80). The method can optionally include a further step in which information is sent from evaluation device 110 to recording device 100 indicating that an entry for the license plate has been found in the booking database (not shown).

[0165] A further embodiment of the method is described in Fig. 13 which does not fall within the scope of the claims.

[0166] The camera 201 of the detection device 200 takes a photo which includes a license plate number of a motor vehicle (step 10).

[0167] Using the processor 202 of the capture device 200, an OCR license plate is created based on the photo of the license plate (step 20).

[0168] A vehicle feature of the motor vehicle is detected by the detection unit 205 of the detection device 200 (step 210). In step 210, multiple vehicle features can also be detected. For example, a laser scan of the vehicle can be performed while the vehicle passes the detection device 200 (e.g., passes under or past the detection device 200).

[0169] The further communication unit 204 of the detection device 200 receives (e.g., via short-range radio communication, such as DSRC) vehicle data from an on-board device arranged in the motor vehicle (step 130). The vehicle data includes a reference license plate. The reference license plate is entered manually by a user when setting up the on-board device. When used as intended (i.e., when the vehicle license plate is entered correctly), the reference license plate corresponds to the actual license plate of the motor vehicle. Receipt of the vehicle data can be preceded by a request from the further communication unit 204 to the on-board device, which initiates the transmission of the vehicle data.

[0170] The order of steps 10, 20, 130, and 210 is not important, with the exception that step 20 can only be executed after a photo with the license plate has been taken in step 10. However, the order of the other steps is arbitrary and can be adapted to the respective conditions if necessary. It can also be provided that some of the steps are executed simultaneously. After these four steps have been performed, the photo with the license plate, the OCR license plate, the reference license plate, and at least one vehicle feature are available in the capture device 200 and can be further processed there.

[0171] The OCR identifier is compared with the reference identifier by processor 202 of the capture device 200 (step 140). If the comparison is positive, i.e., the OCR identifier and the reference identifier match, the method is completed (step 141).

[0172] If the OCR license plate and the reference license plate do not match, the OCR license plate, the reference license plate and the vehicle feature (or vehicle features) are transmitted to the communication unit 112 of the evaluation device 110 using the communication unit 203 of the detection device 200 (step 225).

[0173] A hash value calculation is carried out by the processor 111 of the evaluation device 110 (step 230). The hash value calculation includes the options already described in connection with Fig. 12 were described.

[0174] Alternatively, the hash value calculation can be carried out by means of the processor 202 of the detection device 200, in which case the calculated hash value or the calculated hash values ​​are transmitted from the detection device 200 to the evaluation device 110 instead of the OCR license plate, and possibly also instead of the vehicle characteristics, insofar as these vehicle characteristics were part of the hash value calculation.

[0175] In the next step, the result of the hash value calculation, ie the hash value or hash values, is compared with a correction database using the processor 111 of the evaluation device 110 (step 240). The correction database is structured in accordance with the hash value calculation, as also already described in connection with Fig. 12 described.

[0176] If the comparison with the correction database does not produce a match, or if the comparison confirms the correctness of the OCR license plate, the OCR license plate is forwarded for user-assisted further processing (step 51). If the OCR license plate is correct and the license plate does not match the reference license plate, an attempt at toll fraud may have occurred, which will be investigated. User-assisted further processing may include sending the photo with the license plate from the capture device 200 to the evaluation device 110.

[0177] If the comparison is positive and a correction instruction (or several correction instructions) for the OCR identifier are present, the correction instruction (or the correction instructions) are applied to the OCR identifier using the processor 111 of the evaluation device 100 and a corrected OCR identifier is generated (step 60).

[0178] The corrected OCR identifier is compared with the reference identifier by means of the processor 111 of the evaluation device 110 (step 175).

[0179] In the alternative case that the hash value calculation is performed by the processor 202 of the capture device 200, this comparison is performed by the processor 202 of the capture device 200, which has previously received the corrected OCR identifier from the evaluation device 110.

[0180] If the comparison is positive, i.e., the corrected OCR license plate and the reference license plate match, the method is completed (step 171). It is assumed that the vehicle's on-board device is correctly configured and the vehicle is participating in the toll collection process. The method can optionally include a step in which the evaluation device 110 transmits information about the positive result of the comparison to the detection device 200.

[0181] If the corrected OCR license plate and the reference license plate do not match, the process proceeds to user-assisted further processing (step 180). User-assisted further processing may include sending the photo of the license plate from the capture device 200 to the evaluation device 110. The method may optionally include a step in which the evaluation device 110 transmits information about the negative result of the comparison to the capture device 200.

[0182] After step 20, the method may optionally include calculating one or more confidence values ​​for the text recognition of the license plate using the processor 202 of the detection device 200. In the case of user-assisted further processing, the confidence value (or confidence values) may be an indication of the reliability of the text recognition.

[0183] In an alternative embodiment, in step 155, the OCR license plate, the reference license plate, and the vehicle feature (or vehicle features) are transmitted from the capture device 200 to the evaluation device 110 together with the photo of the license plate. In this case, transmitting the photo is not required during user-assisted further processing (steps 51 or 171).

[0184] The embodiments of the method disclosed here comprise steps of user-controlled further processing. In a first case, user-controlled further processing is required if the comparison of the hash value (or the hash value and the vehicle feature(s)) with the correction database does not produce a match. One reason for this may be that the license plate number was captured for the first time and incorrectly recognized by the text recognition system. In this case, there is no entry in the correction database yet. A second case for user-controlled further processing occurs if the comparison of the corrected OCR license plate number with the booking database is negative. In these cases, an attempt at toll fraud or a technical defect may be present.

[0185] During user-assisted further processing (or verification), a user checks the available data. The user compares the license plate number shown in the photo with the OCR license plate number. If the text recognition is correct, i.e., the OCR license plate number matches the license plate number shown in the photo, this may be a case of toll evasion or a technical malfunction. The user then initiates the next steps. This case is not the subject of the patent application and will not be considered further.

[0186] If the user determines that the OCR of the license plate was unsuccessful, meaning there are discrepancies between the license plate in the photo and the OCR license plate, the user corrects the OCR license plate. An example of a possible OCR error is recognizing the license plate number "B TC 123" as "B TO 123," meaning the "C" is incorrectly recognized as "O." The corrected OCR license plate can then be compared with the booking database.

[0187] The user then creates a correction instruction for the license plate. The hash value for the (incorrectly recognized) OCR license plate is stored in a correction database along with at least one correction instruction (in the above example: replace "O" with "C" or replace the character in position 4 with "C" (spaces are counted)). Thus, the license plate itself is not stored, which could be problematic under data protection law, but only its hash value, from which the license plate cannot be reconstructed. When the license plate is subsequently checked, the OCR license plate can then be automatically corrected according to the stored correction instruction, without the user having to manually intervene again.

[0188] In another variant of user-assisted processing, the photo with the license plate number, along with the OCR license plate number and one or more vehicle characteristics, can be transmitted as a defect data record to a defect database in a data storage of a defect processing system. The user can access the defect data record using operating means of the defect processing system to visually display the defect data record using a display device of the defect processing system. Using the operating means, the user can add a correct license plate number in text form ("text license plate number") as an additional defect data element to a defect data record after visually checking the defect data record.A processor of the error processing system can be configured to generate a correction instruction (or multiple correction instructions) upon receipt of a text identifier in the error database by comparing the text identifier with the OCR identifier. The correction instruction describes how to generate a corrected OCR identifier from the OCR identifier that matches the text identifier if the text identifier and the OCR identifier are different from each other, or to generate a correctness confirmation if the text identifier and the OCR identifier are identical. The processor of the error processing system can then generate a correction data record containing the hash value of the (incorrect or correct) OCR identifier, the correction instruction or the correctness confirmation, and the vehicle characteristic (or characteristics), and initiate storage of the correction data record in the correction database.

Claims

1. A computer-implemented method for correcting a licence plate of a motor vehicle, comprising the following steps: a) capturing (10) a licence plate, b) generating (20) an OCR licence plate from the captured licence plate by means of text recognition, c) performing a reference test for the generated OCR licence plate, wherein the reference test involves comparing the OCR licence plate with a booking database (30), the booking database containing stored licence plates, each associated with vehicles that have already paid a toll fee for a toll road or have booked a toll road, and wherein the reference test is negative if the comparison is negative, with the following steps being executed only if the reference test is negative, d) creating (40) a hash value based on the OCR licence plate generated, e) comparing (50) the created hash value with a correction database, wherein the correction database contains stored hash values of licence plates and correction instructions, with at least one correction instruction assigned to each stored hash value, and f) if the created hash value matches a hash value stored in the correction database, applying (60) the at least one correction instruction assigned to the stored hash value to the OCR licence plate and generating a corrected OCR licence plate.

2. The method according to claim 1, wherein the OCR licence plate comprises one or more characters and wherein the at least one correction instruction has one of the following forms: - replace character "X" with character "Y", - replace the character at position n with character "Y" or - replace character "X" at position n with character "Y", wherein "X" and "Y" are individual characters of a character set and wherein n is a natural number indicating the position of the character in the OCR licence plate.

3. The method according to claim 1 or 2, further comprising capturing (210) at least one vehicle characteristic, wherein in the correction database each stored hash value is associated with information on at least one vehicle characteristic and wherein for comparison with the correction database, the generated hash value and the at least one captured vehicle characteristic are compared with the hash values stored in the correction database and the respective associated information on at least one vehicle characteristic.

4. The method according to claim 3, wherein the at least one vehicle characteristic is selected from the following group: vehicle type, number of axles of a towing vehicle, vehicle height, vehicle width, vehicle length, colour of the vehicle and position of the licence plate on the vehicle.

5. A system for correcting a licence plate of a motor vehicle, wherein the system is configured to perform the following steps: a) capturing (10) a licence plate, b) generating (20) an OCR licence plate from the captured licence plate by means of text recognition, c) performing a reference test for the generated OCR licence plate, wherein the reference test involves comparing the OCR licence plate with a booking database (30), the booking database containing stored licence plates, each associated with vehicles that have already paid a toll fee for a toll road or have booked a toll road, and wherein the reference test is negative if the comparison is negative, with the following steps being executed only if the reference test is negative, d) creating (40) a hash value based on the OCR licence plate generated, e) comparing (50) the created hash value with a correction database, wherein the correction database contains stored hash values of licence plates and correction instructions, with at least one correction instruction assigned to each stored hash value, and f) if the created hash value matches a hash value stored in the correction database, applying (60) the at least one correction instruction assigned to the stored hash value to the OCR licence plate and generating a corrected OCR licence plate.