WO2019066641A2 - A system and method for recognising license plate - Google Patents

A system and method for recognising license plate Download PDF

Info

Publication number
WO2019066641A2
WO2019066641A2 PCT/MY2018/050063 MY2018050063W WO2019066641A2 WO 2019066641 A2 WO2019066641 A2 WO 2019066641A2 MY 2018050063 W MY2018050063 W MY 2018050063W WO 2019066641 A2 WO2019066641 A2 WO 2019066641A2
Authority
WO
WIPO (PCT)
Prior art keywords
license plates
filtered
plate
characters
sub
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/MY2018/050063
Other languages
French (fr)
Other versions
WO2019066641A3 (en
Inventor
Kim Meng Liang
Hamam MOKAYED
Yan Chai HUM
Hock Woon Hon
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Mimos Bhd
Original Assignee
Mimos Bhd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Mimos Bhd filed Critical Mimos Bhd
Publication of WO2019066641A2 publication Critical patent/WO2019066641A2/en
Publication of WO2019066641A3 publication Critical patent/WO2019066641A3/en
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/243—Classification techniques relating to the number of classes
    • G06F18/24323—Tree-organised classifiers
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/60—Type of objects
    • G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
    • G06V20/625—License plates

Definitions

  • the present invention relates to a system and method for recognising license plate. More particularly, the present invention relates to a system and method for recognising a plurality of license plates by using multi-bag analysis.
  • identity of each vehicle is obtained by automatically detecting location of license plate of the vehicle, segmenting the characters out from the detected license plate and recognising the segmented characters into a readable text format. Therefore the identity of the vehicle is determined from the readable text format of the recognised license plate.
  • a United States Patent Publication No. 2013/0259314 A1 discloses an example of a method and system for enhancing performance of an automated license plate recognition.
  • the automated license plate recognition system utilises multiple results by capturing multiple images as a vehicle passes through an observation zone and processing each image to obtain plate code results and associated confidence values. Character level information is iteratively processed and optical recognition code is optimally aligned with respect to the multiple images by utilising a string correlation approach to obtain the result.
  • the present invention relates to a system (100) and method for recognising license plate.
  • the system (100) comprising an image acquisition module (10) configured to obtain a plurality of continuous frames, a plate detection module (20) configured to detect the location of the license plate inside the plurality of continuous frames, and a plate recognition module (30) configured to recognise and transform contents of the plate image received from the plate detection module (20) into character text format.
  • the system (100) further comprising a post- processing module (40) configured to analyse a plurality of recognised license plates from the plate recognition module (30) using multi-bag analysis to derive a final recognition result from the plurality of recognised license plates by classifying the plurality of recognised license plates based on the length of the recognised license plates and number of identical numeric and alphabet of the recognised license plates.
  • the post-processing module (40) comprising a non-plate filter (41) configured to filter noises that resemble license plate from the plurality of recognised license plates based on the number of alphabet and numeric characters a big bag creation and selection sub-module (42) configured to group of the plurality of filtered license plates based on identical number of alphabet and numeric characters, wherein the plurality of filtered license plates refer to the recognised license plates that are not discarded by the non-plate filter (41), a mini bag creation and selection sub-module (43) configured to determine the most related sub-group in the group based on character length of the plurality of filtered license plate, an adaptive frequency updater (44) configured to update frequency value for each characters location of the plurality of filtered license plates in each sub-group, and an adaptive recognition analyser (45) configured to determine the final recognition result by analysing the updated frequency values in each sub-group.
  • a non-plate filter (41) configured to filter noises that resemble license plate from the plurality of recognised license plates based on the number of alphabet and numeric characters
  • the method for recognising license plate includes the steps receiving a plurality of continuous frame of a vehicle passing through a monitoring area by an image acquisition module (10), detecting location of the license plate inside each frame by a plate detection module (20), recognising each character of the license plate of each frame by a plate recognition module (30), filtering noises that resemble license plate from a plurality of recognised license plates based on the number of numeric and alphabet characters by a non-plate filter (41), grouping a plurality of filtered license plates based on the number of characters exists in the plurality of filtered license plates by a big bag creation and selection sub-module (42), wherein the plurality of filtered license plates refer to a plurality of recognised license plates that are not discarded by the non-plate filter (41), determining the most related subgroup for each group based on a character length of the plurality of filtered license plates result by a mini bag creation and selection sub-module (43), updating frequency value for each characters location of the plurality of filtered license plates by an adaptive frequency
  • FIG. 1 illustrates a block diagram of a system (100) for recognising license plate according to an embodiment of the present invention.
  • FIG 2 illustrates a block diagram of sub-modules of a post-processing module (40) of the system (100) of FIG. 1 .
  • FIG. 3 illustrates a flowchart of a method for recognising license plate according to an embodiment of the present invention.
  • FIG. 4 illustrates a flowchart of sub-steps for grouping filtered license plates of the method of FIG. 3.
  • FIG. 5 illustrates a flowchart of sub-steps for updating a frequency value at each characters location of the filtered license plates of the method of FIG. 3.
  • FIG. 6 illustrates a flowchart of sub-steps for determining a final recognition result of the method of FIG. 3.
  • FIG. 1 illustrates a block diagram of a system (100) for recognising license plate according to an embodiment of the present invention.
  • the system (100) comprises an image acquisition module (10), a plate detection module (20), a plate recognition module (30), a post-processing module (40), and a displaying device (50).
  • the system (100) analyses multiple images that are captured from a moving vehicle which enters a monitoring zone.
  • Final recognition result is obtained by analysing a plurality of recognised license plates from multiple frames using a multi-bag analysis, wherein the multi-bag analysis is a method of classifying the recognised license plates from multiple frames based on the length of the recognised license plates and number of identical numeric and alphabet characters of the recognised license plates.
  • the multi-bag analysis is a method of classifying the recognised license plates from multiple frames based on the length of the recognised license plates and number of identical numeric and alphabet characters of the recognised license plates.
  • the image acquisition module (10) is configured to obtain a plurality of continuous frames from a video capturing device.
  • the plate detection module (20) is connected to the image acquisition module (10) and the plate recognition module (30).
  • the frames obtained by the image acquisition module (10) are processed by the plate detection module (20) to detect the location of the license plate inside the frames.
  • the detected license plates are sent to the plate recognition module (30).
  • the plate recognition module (30) is connected to the plate detection module (20) and the post-processing module (40).
  • the plate recognition module (30) is configured to recognise contents of the license plate image received from the plate detection module (20), transform the contents of the license plate image into character text format and send to the post-processing module (40) for final recognition.
  • the character text format includes both alphabet and numeric.
  • the post-processing module (40) is connected to the plate recognition module (30) and the displaying device (50).
  • the post-processing module (40) is configured to analyse the recognised license plates of every frames from the plate recognition module (30) using multi-bag analysis to derive a final recognition result from the recognised license plates of every frames of each vehicle.
  • the final recognition result is sent to the displaying device (50) to be displayed in real time. As shown in FIG.
  • the post-processing module (40) comprising a non-plate filter (41), a big bag creation and selection sub-module (42), a mini bag creation and selection sub-module (43), an adaptive frequency updater (44) and an adaptive recognition analyser (45).
  • the non-plate filter (41) is configured to filter noises that resemble license plate from the recognised license plates based on the number of alphabet and numeric characters.
  • the non-plate filter (41) receives the recognised license plates of every frames from the plate recognition module (30) and filters out the recognised license plate that has low possibility of a valid plate image.
  • the non-plate filter (41) sends the filtered license plates to the big bag creation and selection sub-module (42).
  • the big bag creation and selection sub-module (42) is configured to group the filtered license plates based on an identical number of alphabet and numeric characters.
  • the mini bag creation and selection sub-module (43) is configured to determine the most related sub-group of each group, wherein sub-groups are subdivisions of the group based on character length.
  • the character length is the total number of alphabet and numeric characters of the filtered license plates.
  • the adaptive frequency updater (44) is configured to update frequency value for each characters location of the filtered license plates in each sub-group, wherein the characters location is a position of either an alphabet or a numeric character in the filtered license plates. Frequency value is total occurrence of a similar character being displayed at the same characters location.
  • the adaptive recognition analyser (45) is configured to determine the final recognition result by analysing the updated frequency values in each sub-group.
  • the displaying device (50) which is connected to the post-analysing module (40) is configured to display the final recognition result of the license plate of the vehicle.
  • FIG. 3 illustrates a flowchart of a method for recognising license plate according to an embodiment of the present invention.
  • the image acquisition module (10) receives a plurality of continuous frames of a vehicle passing through a monitoring area from a video acquiring device as in step 1000.
  • the frames are then sent to the plate detection module (20).
  • the plate detection module (20) then processes the frames and detects the location of the license plate inside each frame as in step 2000.
  • the plate recognition module (30) then recognises each character of the license plate in each frame as in step 3000.
  • the characters include alphabets and numerics.
  • the plate recognition module (30) sends the recognised license plates of every frames to the non-plate filter (41).
  • the non-plate filter (41) filters noises that resemble license plate from the recognised license plates as in step 4000.
  • the noises are the recognised license plate that has a low probability of being a valid license plate.
  • the non-plate filter (41) filters the noises based on the number of numeric and alphabet characters. If the recognised license plate has the number of numeric and alphabet characters less than the minimum value of allowed characters or more than the maximum value of allowed characters, the recognised license plate is discarded by the non-plate filter (41) as noise.
  • the recognised license plates that are not discarded by the non-plate filter (41) are known as the filtered license plates.
  • the filtered license plates are then sent to the big bag creation and selection sub-module (42) for the next step.
  • the big bag creation and selection sub-module (42) receives the filtered license plates from the non-plate filter (41) and groups the filtered license plates as in step 5000.
  • the filtered license plates are grouped based on the number of numeric and alphabet characters exists in the filtered license plates.
  • the sub-steps for grouping the filtered license plates are further explained in relation to FIG. 4.
  • the mini bag creation and selection sub-module (43) determines the most related sub-group for each group based on the character length of the filtered license plates as in step 6000.
  • Sub-groups are subdivisions of the group based on character length, wherein the character length is the total number of alphabet and numeric characters of the filtered license plates.
  • the mini bag creation and selection sub-module (43) computes the character length of every filtered license plates and selects the sub-group that has a similar character length. If none of the sub-groups in the group has a similar character length, the mini bag creation and selection sub- module (42) creates a new sub-group for the filtered license plates.
  • the adaptive frequency updater (44) then updates frequency value for each characters location of the filtered license plates for each sub-group as in step 7000, wherein the characters location is a position of either an alphabet or a numeric character in the filtered license plates. Frequency value is total occurrence of a similar character being displayed for in the same characters location.
  • the sub-steps for updating the frequency value at each characters location of the filtered license plates for each sub-group are further explained in relation to FIG. 5.
  • the adaptive recognition analyser (45) determines the final recognition result as in step 8000.
  • the adaptive recognition analyser (45) computes a voting value for each sub-group to determine whether the sub-group contains the final recognition result that fulfill conditions for publishing the final recognition result.
  • the conditions for publishing the final recognition result are further explained in relation to FIG. 6.
  • the voting value represents the total characters location of the filtered license plates in the sub-group where the frequency value at the characters location had exceeded a certain threshold value.
  • the sub-steps for determining the final recognition result are further explained in relation to FIG. 6.
  • the displaying device (50) displays the final recognition result as in step 9000.
  • FIG. 4 illustrates a flowchart of sub-steps for grouping the filtered license plates of the steps 5000 of the method of FIG. 3.
  • the big bag creation and selection sub-module (41) computes the character length of each filtered license plate as in step 5100.
  • the big bag creation and selection sub-module (41) then computes identical numeric and alphabet characters of each filtered license plate as in step 5200.
  • the big bag creation and selection sub-module (41) removes the identical numeric and alphabet characters from the character length to obtain the number of characters exists in each filtered license plate as in step 5250.
  • the big bag creation and selection sub-module (41) selects a group that best matches each filtered license plate as in step 5300.
  • the group is selected based on the number of characters exists in the filtered license plate. If the number of characters exists in the filtered license plate of the group is the same as the number of characters exists in the filtered license plate, the group is selected.
  • the big bag creation and selection sub-module (41) determines whether an existing group is selected or not as in step 5350. If the existing group is selected, the big bag creation and selection sub-module (41) directly updates properties in the group such as the character length, number of identical numeric, number of identical alphabet and number of characters exists in the filtered license plate as in step 5500. Each group has the properties to assist the process of matching the filtered license plates with the group.
  • the big bag creation and selection (41) creates a new group for the filtered license plate that is not grouped result as in step 5400 and updates the properties in the group as in step 5500.
  • FIG. 5 illustrates a flowchart of sub-steps for updating the frequency value for each characters location of the filtered license plates for each sub-group of step 7000 of the method of FIG. 3.
  • the adaptive frequency updater (44) selects a matched character buffer at characters location of the filtered license plates as in step 7100.
  • the matched character buffer is a character that previously matched with the filtered license plates in the particular sub-group.
  • the adaptive frequency updater (44) determines whether all characters of the filtered license plates match with the matched character buffer at each characters location as in decision 7150.
  • the adaptive frequency updater (44) creates a new matched character buffer at the mismatched characters location of the filtered license plates as in step 7200.
  • the adaptive frequency updater (44) increases a frequency value of the matched character buffer at each characters location of the filtered license plates as in step 7300.
  • the adaptive frequency updater (44) directly increases the frequency value of the matched character buffer at each characters location of the filtered license plates as in step 7300. Thereon, the adaptive frequency updater (44) stores the value of the highest frequency value and its corresponding matched character buffer for each characters location as in step 7400.
  • FIG. 6 illustrates a flowchart of sub-steps for determining the final recognition result of the step 8000 of the method of FIG. 3.
  • the adaptive recognition analyser (45) computes the voting value for each sub-group to determine whether the subgroup contains final recognition result that fulfill the condition for publishing the final recognition result. Initially, the adaptive recognition analyser (45) determines whether the highest frequency value of the filtered license plates in each sub-group exceeds a predefined value as in decision 8050, wherein the predefined value refers to a confidence value of a consistency for a similar character appears on the same characters location.
  • the adaptive recognition analyser (45) rejects the filtered license plates as in step 8800 and no filtered license plates in the sub-group will be published as the final result.
  • the adaptive recognition analyser (45) increases the voting value as in step 8100. Thereon, the adaptive recognition analyser (45) determines whether the voting value is equal to the character length of the filtered license plates as in decision 8150.
  • the adaptive recognition analyser (45) increases a delay timer as in step 8200, wherein the delay timer is a timer for a sub-group with voting value that is more than the predefined value. Thereon, the adaptive recognition analyser (45) determines whether the delay timer is greater than a threshold value of delay timer as in decision 8250. If the delay timer is less than the threshold value of delay timer as in decision 8200, the adaptive recognition analyser (45) rejects the filtered license plates as in step 8800 and no filtered license plates in the sub-group will be published as the final result. The system (100) waits for a later recognition which may be more accurate, wherein the later recognition result is the filtered license plates that are obtained from frames after the frames of the filtered license plates.
  • the adaptive recognition analyser (45) published the characters with the highest frequency as the final recognition result as in step 8300.
  • the adaptive recognition analyser (45) determines whether the voting value is greater than the minimum number of character as in decision 8400. If the voting value is less than or equals to the minimum number of character, the adaptive recognition analyser (45) rejects the filtered license plates as in step 8800 because the filtered license plates does not meet the confidence value of the system (100) and may be noises that were not filtered by the non-plate filter (41). Additionally, no filtered license plates in the sub-group will be published as the final result.
  • the adaptive recognition analyser (45) increases a silence timer as in step 8500.
  • the silence timer is a timer for a sub-group that does not have voting value more than a predefined value but has more than a minimum voting value to ensure there is a final recognition result to be published.
  • the adaptive recognition analyser (45) determines whether the silence timer is greater than a threshold value of silence timer as in decision 8600. If the silence timer is less than the threshold value of silence timer, the adaptive recognition analyser (45) rejects the filtered license plates as in step 8800 and no filtered license plates in the sub-group will be published as the final result. The system (100) waits for the later recognition result. If the silence timer is greater than the threshold value of silence timer, the adaptive recognition analyser (45) determines whether the number of filtered license plates in the sub-group is greater than a minimum number of filtered license plates in the sub-group as in decision 8700. If the number of filtered license plates in the sub- group is greater than a minimum number of filtered license plates in the sub-group, the adaptive recognition analyser (45) published the characters with the highest frequency as the final recognition result as in step 8300.
  • the adaptive recognition analyser rejects the filtered license plates as in step 8800 and no filtered license plates in the sub-group will be published as the final result.

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Artificial Intelligence (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Traffic Control Systems (AREA)
  • Character Discrimination (AREA)
  • Character Input (AREA)

Abstract

The present invention relates to a system (100) and method for recognising license plate. The system (100) comprising an image acquisition module (10) configured to obtain a plurality of continuous frames, a plate detection module (20) configured to detect the location of the license plate inside the plurality of continuous frames, and a plate recognition module (30) configured to recognise and transform content of the license plate image received from the plate detection module (20) into character text format. The system (100) further comprising a post-processing module (40) configured to analyse a plurality of recognised license plates from the plate recognition module (30) using multi-bag analysis to derive a final recognition result from the plurality of recognised license plates by classifying the plurality of recognised license plates based on the length of the recognised license plates and number of identical numeric and alphabet of the recognised license plates.

Description

A SYSTEM AND METHOD FOR RECOGNISING LICENSE PLATE
FIELD OF INVENTION
The present invention relates to a system and method for recognising license plate. More particularly, the present invention relates to a system and method for recognising a plurality of license plates by using multi-bag analysis.
BACKGROUND OF THE INVENTION
In an automated license plate recognition system, identity of each vehicle is obtained by automatically detecting location of license plate of the vehicle, segmenting the characters out from the detected license plate and recognising the segmented characters into a readable text format. Therefore the identity of the vehicle is determined from the readable text format of the recognised license plate.
A United States Patent Publication No. 2013/0259314 A1 discloses an example of a method and system for enhancing performance of an automated license plate recognition. The automated license plate recognition system utilises multiple results by capturing multiple images as a vehicle passes through an observation zone and processing each image to obtain plate code results and associated confidence values. Character level information is iteratively processed and optical recognition code is optimally aligned with respect to the multiple images by utilising a string correlation approach to obtain the result.
Although there are many systems and methods for automatically recognising a license plate, most of the systems and methods that compute final recognition result from multiple images have a tendency to display the final recognition result only after the vehicle has left the monitoring zone. This is proven to be a disadvantage as the identity of the vehicle may fall in a black-listed category, whereby an immediate action is needed. Therefore, there is a need for a system and method for automatically recognising a license plate that can improve the processing time of the system for automatically recognising a license plate.
SUMMARY OF INVENTION
The present invention relates to a system (100) and method for recognising license plate. The system (100) comprising an image acquisition module (10) configured to obtain a plurality of continuous frames, a plate detection module (20) configured to detect the location of the license plate inside the plurality of continuous frames, and a plate recognition module (30) configured to recognise and transform contents of the plate image received from the plate detection module (20) into character text format. The system (100) further comprising a post- processing module (40) configured to analyse a plurality of recognised license plates from the plate recognition module (30) using multi-bag analysis to derive a final recognition result from the plurality of recognised license plates by classifying the plurality of recognised license plates based on the length of the recognised license plates and number of identical numeric and alphabet of the recognised license plates.
Preferably, the post-processing module (40) comprising a non-plate filter (41) configured to filter noises that resemble license plate from the plurality of recognised license plates based on the number of alphabet and numeric characters a big bag creation and selection sub-module (42) configured to group of the plurality of filtered license plates based on identical number of alphabet and numeric characters, wherein the plurality of filtered license plates refer to the recognised license plates that are not discarded by the non-plate filter (41), a mini bag creation and selection sub-module (43) configured to determine the most related sub-group in the group based on character length of the plurality of filtered license plate, an adaptive frequency updater (44) configured to update frequency value for each characters location of the plurality of filtered license plates in each sub-group, and an adaptive recognition analyser (45) configured to determine the final recognition result by analysing the updated frequency values in each sub-group.
The method for recognising license plate includes the steps receiving a plurality of continuous frame of a vehicle passing through a monitoring area by an image acquisition module (10), detecting location of the license plate inside each frame by a plate detection module (20), recognising each character of the license plate of each frame by a plate recognition module (30), filtering noises that resemble license plate from a plurality of recognised license plates based on the number of numeric and alphabet characters by a non-plate filter (41), grouping a plurality of filtered license plates based on the number of characters exists in the plurality of filtered license plates by a big bag creation and selection sub-module (42), wherein the plurality of filtered license plates refer to a plurality of recognised license plates that are not discarded by the non-plate filter (41), determining the most related subgroup for each group based on a character length of the plurality of filtered license plates result by a mini bag creation and selection sub-module (43), updating frequency value for each characters location of the plurality of filtered license plates by an adaptive frequency updater (44), and determining a final recognition result by an adaptive recognition analyser (45).
BRIEF DESCRIPTION OF THE DRAWINGS
The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the invention and, together with the description, serve to explain the principles of the invention.
FIG. 1 illustrates a block diagram of a system (100) for recognising license plate according to an embodiment of the present invention.
FIG 2 illustrates a block diagram of sub-modules of a post-processing module (40) of the system (100) of FIG. 1 .
FIG. 3 illustrates a flowchart of a method for recognising license plate according to an embodiment of the present invention.
FIG. 4 illustrates a flowchart of sub-steps for grouping filtered license plates of the method of FIG. 3. FIG. 5 illustrates a flowchart of sub-steps for updating a frequency value at each characters location of the filtered license plates of the method of FIG. 3.
FIG. 6 illustrates a flowchart of sub-steps for determining a final recognition result of the method of FIG. 3.
DESCRIPTION OF THE PREFERRED EMBODIMENT
A preferred embodiment of the present invention will be described herein below with reference to the accompanying drawings. In the following description, well known functions or constructions are not described in detail since they would obscure the description with unnecessary detail. Reference is initially made to FIG. 1 which illustrates a block diagram of a system (100) for recognising license plate according to an embodiment of the present invention. The system (100) comprises an image acquisition module (10), a plate detection module (20), a plate recognition module (30), a post-processing module (40), and a displaying device (50). Generally, the system (100) analyses multiple images that are captured from a moving vehicle which enters a monitoring zone. Final recognition result is obtained by analysing a plurality of recognised license plates from multiple frames using a multi-bag analysis, wherein the multi-bag analysis is a method of classifying the recognised license plates from multiple frames based on the length of the recognised license plates and number of identical numeric and alphabet characters of the recognised license plates.
The image acquisition module (10) is configured to obtain a plurality of continuous frames from a video capturing device.
The plate detection module (20) is connected to the image acquisition module (10) and the plate recognition module (30). The frames obtained by the image acquisition module (10) are processed by the plate detection module (20) to detect the location of the license plate inside the frames. The detected license plates are sent to the plate recognition module (30).
The plate recognition module (30) is connected to the plate detection module (20) and the post-processing module (40). The plate recognition module (30) is configured to recognise contents of the license plate image received from the plate detection module (20), transform the contents of the license plate image into character text format and send to the post-processing module (40) for final recognition. The character text format includes both alphabet and numeric. The post-processing module (40) is connected to the plate recognition module (30) and the displaying device (50). The post-processing module (40) is configured to analyse the recognised license plates of every frames from the plate recognition module (30) using multi-bag analysis to derive a final recognition result from the recognised license plates of every frames of each vehicle. The final recognition result is sent to the displaying device (50) to be displayed in real time. As shown in FIG. 2 which illustrates a block diagram of sub-modules of the postprocessing module (40) of the system (100), the post-processing module (40) comprising a non-plate filter (41), a big bag creation and selection sub-module (42), a mini bag creation and selection sub-module (43), an adaptive frequency updater (44) and an adaptive recognition analyser (45).
The non-plate filter (41) is configured to filter noises that resemble license plate from the recognised license plates based on the number of alphabet and numeric characters. The non-plate filter (41) receives the recognised license plates of every frames from the plate recognition module (30) and filters out the recognised license plate that has low possibility of a valid plate image. The non-plate filter (41) sends the filtered license plates to the big bag creation and selection sub-module (42). The big bag creation and selection sub-module (42) is configured to group the filtered license plates based on an identical number of alphabet and numeric characters. The mini bag creation and selection sub-module (43) is configured to determine the most related sub-group of each group, wherein sub-groups are subdivisions of the group based on character length. The character length is the total number of alphabet and numeric characters of the filtered license plates. The adaptive frequency updater (44) is configured to update frequency value for each characters location of the filtered license plates in each sub-group, wherein the characters location is a position of either an alphabet or a numeric character in the filtered license plates. Frequency value is total occurrence of a similar character being displayed at the same characters location. The adaptive recognition analyser (45) is configured to determine the final recognition result by analysing the updated frequency values in each sub-group.
Referring back to FIG. 1 , the displaying device (50) which is connected to the post-analysing module (40) is configured to display the final recognition result of the license plate of the vehicle.
Reference is now made to FIG. 3 which illustrates a flowchart of a method for recognising license plate according to an embodiment of the present invention. Initially, the image acquisition module (10) receives a plurality of continuous frames of a vehicle passing through a monitoring area from a video acquiring device as in step 1000. The frames are then sent to the plate detection module (20). The plate detection module (20) then processes the frames and detects the location of the license plate inside each frame as in step 2000.
The plate recognition module (30) then recognises each character of the license plate in each frame as in step 3000. The characters include alphabets and numerics. The plate recognition module (30) sends the recognised license plates of every frames to the non-plate filter (41).
Thereon, the non-plate filter (41) filters noises that resemble license plate from the recognised license plates as in step 4000. The noises are the recognised license plate that has a low probability of being a valid license plate. The non-plate filter (41) filters the noises based on the number of numeric and alphabet characters. If the recognised license plate has the number of numeric and alphabet characters less than the minimum value of allowed characters or more than the maximum value of allowed characters, the recognised license plate is discarded by the non-plate filter (41) as noise. The recognised license plates that are not discarded by the non-plate filter (41) are known as the filtered license plates. The filtered license plates are then sent to the big bag creation and selection sub-module (42) for the next step.
The big bag creation and selection sub-module (42) receives the filtered license plates from the non-plate filter (41) and groups the filtered license plates as in step 5000. The filtered license plates are grouped based on the number of numeric and alphabet characters exists in the filtered license plates. The sub-steps for grouping the filtered license plates are further explained in relation to FIG. 4.
Thereon, the mini bag creation and selection sub-module (43) determines the most related sub-group for each group based on the character length of the filtered license plates as in step 6000. Sub-groups are subdivisions of the group based on character length, wherein the character length is the total number of alphabet and numeric characters of the filtered license plates. The mini bag creation and selection sub-module (43) computes the character length of every filtered license plates and selects the sub-group that has a similar character length. If none of the sub-groups in the group has a similar character length, the mini bag creation and selection sub- module (42) creates a new sub-group for the filtered license plates.
The adaptive frequency updater (44) then updates frequency value for each characters location of the filtered license plates for each sub-group as in step 7000, wherein the characters location is a position of either an alphabet or a numeric character in the filtered license plates. Frequency value is total occurrence of a similar character being displayed for in the same characters location. The sub-steps for updating the frequency value at each characters location of the filtered license plates for each sub-group are further explained in relation to FIG. 5.
The adaptive recognition analyser (45) determines the final recognition result as in step 8000. The adaptive recognition analyser (45) computes a voting value for each sub-group to determine whether the sub-group contains the final recognition result that fulfill conditions for publishing the final recognition result. The conditions for publishing the final recognition result are further explained in relation to FIG. 6. The voting value represents the total characters location of the filtered license plates in the sub-group where the frequency value at the characters location had exceeded a certain threshold value. The sub-steps for determining the final recognition result are further explained in relation to FIG. 6. Finally, the displaying device (50) displays the final recognition result as in step 9000.
FIG. 4 illustrates a flowchart of sub-steps for grouping the filtered license plates of the steps 5000 of the method of FIG. 3. Initially, the big bag creation and selection sub-module (41) computes the character length of each filtered license plate as in step 5100. The big bag creation and selection sub-module (41) then computes identical numeric and alphabet characters of each filtered license plate as in step 5200. The big bag creation and selection sub-module (41) removes the identical numeric and alphabet characters from the character length to obtain the number of characters exists in each filtered license plate as in step 5250.
Thereon, the big bag creation and selection sub-module (41) selects a group that best matches each filtered license plate as in step 5300. The group is selected based on the number of characters exists in the filtered license plate. If the number of characters exists in the filtered license plate of the group is the same as the number of characters exists in the filtered license plate, the group is selected.
The big bag creation and selection sub-module (41) determines whether an existing group is selected or not as in step 5350. If the existing group is selected, the big bag creation and selection sub-module (41) directly updates properties in the group such as the character length, number of identical numeric, number of identical alphabet and number of characters exists in the filtered license plate as in step 5500. Each group has the properties to assist the process of matching the filtered license plates with the group.
On the other hand, if none of the existing groups is selected, the big bag creation and selection (41) creates a new group for the filtered license plate that is not grouped result as in step 5400 and updates the properties in the group as in step 5500.
FIG. 5 illustrates a flowchart of sub-steps for updating the frequency value for each characters location of the filtered license plates for each sub-group of step 7000 of the method of FIG. 3. After the most suitable sub-group is selected by the mini bag creation and selection sub-module (43) as in step 6000, the adaptive frequency updater (44) selects a matched character buffer at characters location of the filtered license plates as in step 7100. The matched character buffer is a character that previously matched with the filtered license plates in the particular sub-group. The adaptive frequency updater (44) then determines whether all characters of the filtered license plates match with the matched character buffer at each characters location as in decision 7150. If there is at least one character that does not match with the matched character buffer, the adaptive frequency updater (44) creates a new matched character buffer at the mismatched characters location of the filtered license plates as in step 7200. The adaptive frequency updater (44) increases a frequency value of the matched character buffer at each characters location of the filtered license plates as in step 7300.
On the other hand, if all the characters of the filtered license plates match with the matched character at each characters location as in decision 7150, the adaptive frequency updater (44) directly increases the frequency value of the matched character buffer at each characters location of the filtered license plates as in step 7300. Thereon, the adaptive frequency updater (44) stores the value of the highest frequency value and its corresponding matched character buffer for each characters location as in step 7400.
FIG. 6 illustrates a flowchart of sub-steps for determining the final recognition result of the step 8000 of the method of FIG. 3. The adaptive recognition analyser (45) computes the voting value for each sub-group to determine whether the subgroup contains final recognition result that fulfill the condition for publishing the final recognition result. Initially, the adaptive recognition analyser (45) determines whether the highest frequency value of the filtered license plates in each sub-group exceeds a predefined value as in decision 8050, wherein the predefined value refers to a confidence value of a consistency for a similar character appears on the same characters location.
If the highest frequency value of the filtered license plates in the sub-group does not exceed the confidence value of a consistency for a similar character appears on the same characters location, the adaptive recognition analyser (45) rejects the filtered license plates as in step 8800 and no filtered license plates in the sub-group will be published as the final result. On the other hand, if the highest frequency value of the filtered license plates in the sub-group exceeds the confidence value of a consistency for a similar character appears on the same characters location, the adaptive recognition analyser (45) increases the voting value as in step 8100. Thereon, the adaptive recognition analyser (45) determines whether the voting value is equal to the character length of the filtered license plates as in decision 8150.
If the voting value is equal to the character length of the filtered license plates, the adaptive recognition analyser (45) increases a delay timer as in step 8200, wherein the delay timer is a timer for a sub-group with voting value that is more than the predefined value. Thereon, the adaptive recognition analyser (45) determines whether the delay timer is greater than a threshold value of delay timer as in decision 8250. If the delay timer is less than the threshold value of delay timer as in decision 8200, the adaptive recognition analyser (45) rejects the filtered license plates as in step 8800 and no filtered license plates in the sub-group will be published as the final result. The system (100) waits for a later recognition which may be more accurate, wherein the later recognition result is the filtered license plates that are obtained from frames after the frames of the filtered license plates.
On the other hand, if the delay timer is greater than the threshold value of delay timer, the adaptive recognition analyser (45) published the characters with the highest frequency as the final recognition result as in step 8300.
If the voting value is not equal to the character length as in decision 8150, the adaptive recognition analyser (45) determines whether the voting value is greater than the minimum number of character as in decision 8400. If the voting value is less than or equals to the minimum number of character, the adaptive recognition analyser (45) rejects the filtered license plates as in step 8800 because the filtered license plates does not meet the confidence value of the system (100) and may be noises that were not filtered by the non-plate filter (41). Additionally, no filtered license plates in the sub-group will be published as the final result.
Otherwise, if the voting value is greater than the minimum number of character as in decision 8400, the adaptive recognition analyser (45) increases a silence timer as in step 8500. The silence timer is a timer for a sub-group that does not have voting value more than a predefined value but has more than a minimum voting value to ensure there is a final recognition result to be published.
The adaptive recognition analyser (45) then determines whether the silence timer is greater than a threshold value of silence timer as in decision 8600. If the silence timer is less than the threshold value of silence timer, the adaptive recognition analyser (45) rejects the filtered license plates as in step 8800 and no filtered license plates in the sub-group will be published as the final result. The system (100) waits for the later recognition result. If the silence timer is greater than the threshold value of silence timer, the adaptive recognition analyser (45) determines whether the number of filtered license plates in the sub-group is greater than a minimum number of filtered license plates in the sub-group as in decision 8700. If the number of filtered license plates in the sub- group is greater than a minimum number of filtered license plates in the sub-group, the adaptive recognition analyser (45) published the characters with the highest frequency as the final recognition result as in step 8300.
On the other hand, if the silence timer is less than or equals to the threshold value of the silence timer as in decision 8700, the adaptive recognition analyser (45) rejects the filtered license plates as in step 8800 and no filtered license plates in the sub-group will be published as the final result.
While embodiments of the invention have been illustrated and described, it is not intended that these embodiments illustrate and describe all possible forms of the invention. Rather, the words used in the specifications are words of description rather than limitation and various changes may be made without departing from the scope of the invention.

Claims

1 . A system (100) for recognising license plate comprising:
a) an image acquisition module (10) configured to obtain a plurality of continuous frames;
b) a plate detection module (20) configured to detect the location of the license plate image inside the plurality of continuous frames; and c) a plate recognition module (30) configured to recognise and transform contents of the license plate image received from the plate detection module (20) into character text format;
characterised in that the system (100) further comprising:
d) a post-processing module (40) configured to analyse a plurality of recognised license plates from the plate recognition module (30) using multi-bag analysis to derive a final recognition result from the plurality of recognised license plates by classifying the plurality of recognised license plates based on the length of the recognised license plates and number of identical numeric and alphabet of the recognised license plates.
2. The system (100) as claimed in claim 1 , wherein the post-processing module (40) comprising:
a) a non-plate filter (41) configured to filter noises that resemble license plate from the plurality of recognised licence plates based on the number of alphabet and numeric characters;
b) a big bag creation and selection sub-module (42) configured to group a plurality of filtered license plates based on identical number of alphabet and numeric characters, wherein the plurality of filtered license plates refers to the recognised license plates that are not discarded by the non-plate filter (41);
c) a mini bag creation and selection sub-module (43) configured to determine the most related sub-group in the group based on character length of the plurality of filtered license plate;
d) an adaptive frequency updater (44) configured to update frequency value for each characters location of the plurality of filtered license plates in each sub-group; and e) an adaptive recognition analyser (45) configured to determine the final recognition result by analysing the updated frequency values in each sub-group.
The system (100) as claimed in claim 1 , wherein the system (100) further comprising a displaying device (50) configured to display the final recognition result of the license plate of the vehicle.
A method for recognising license plate is characterised by the steps of:
a) receiving a plurality of continuous frame of a vehicle passing through a monitoring area by an image acquisition module (10);
b) detecting location of the license plate inside each frame by a plate detection module (20);
c) recognising each character of the license plate of each frame by a plate recognition module (30);
d) filtering noises that resemble license plate from a plurality of recognised license plates based on the number of numeric and alphabet characters by a non-plate filter (41);
e) grouping a plurality of filtered license plates based on the number of characters exists in the plurality of filtered license plates by a big bag creation and selection sub-module (42), wherein the plurality of filtered license plates is the plurality of recognised license plates that are not discarded by the non-plate filter (41);
f) determining the most related sub-group for each group based on a character length of the plurality of filtered license plates result by a mini bag creation and selection sub-module (43);
g) updating frequency value for each characters location of the plurality of filtered license plates by an adaptive frequency updater (44); and h) determining a final recognition result by an adaptive recognition analyser (45).
The method as claimed in claim 4, wherein grouping the plurality of filtered license plates based on the number of characters exists in the plurality of filtered license plates includes the steps of:
a) computing character length of each filtered license plate; b) computing identical numeric and alphabet characters of each filtered license plate;
c) obtaining the number of characters exists in each filtered license plate; d) selecting an existing group that best matches each filtered license plate based on the number of characters exists in the filtered license plate;
e) determining whether an existing group is selected or not; and f) creating a new group for the filtered license plate if none of the existing groups is selected.
The method as claimed in claim 4, wherein determining the most related subgroup for each group based on the character length of the plurality of filtered license plates by the mini bag creation and selection sub-module (43) includes the steps of:
a) computing character length of each filtered license plate;
b) selecting an existing sub-group that has similar character length as the filtered license plate;
c) determining whether an existing sub-group is selected or not; and d) creating a new group for the filtered license plate if none of the existing sub-groups is selected.
The method as claimed in claim 4, wherein updating frequency value for each characters location of the plurality of filtered license plates includes the steps of:
a) selecting a matched character buffer at each characters location of the plurality of filtered license plates;
b) determining whether all characters of the plurality of filtered license plates match with the matched character buffer at each characters location;
c) creating a new matched character buffer at the mismatched characters location of the plurality of filtered license plates if there is at least one character that does not match the matched character buffer; and
d) increasing a frequency value of the matched character buffer at each characters location of the recognition.
The method as claimed in claim 7, wherein if all the characters of the plurality of filtered license plates match with the matched character buffer of each character location, the step includes increasing a frequency value of the matched character buffer at each characters location of the plurality of filtered license plates.
The method as claimed in claim 4, wherein the step of determining the final recognition result by the adaptive recognition analyser (45) includes the sub- steps of:
a) determining whether the highest frequency value of the plurality of filtered license plates in each sub-group exceeds a confidence value of a consistency for a similar character appears on the same characters location;
b) increasing a voting value if the highest frequency value of the plurality of filtered license plates in the sub-group exceeds the confidence value of a consistency for a similar character appears on the same characters location;
c) determining whether the voting value is equal to a character length of the plurality of filtered license plates;
d) increasing a delay timer if the voting value is equal to the character length of the plurality of filtered license plates;
e) determining whether the delay timer is greater than a threshold value of delay timer; and
f) publishing the characters with the highest frequency as the final recognition result if the delay timer is greater than the threshold value of delay timer.
The method as claimed in claim 9, wherein the step includes rejecting the plurality of filtered license plates, if the highest frequency value of the plurality of filtered license plates in the sub-group does not exceed the confidence value of how consistent a similar character appears on the same characters location.
The method as claimed in claim 9, wherein if the voting value is not equal to the character length, the steps include:
a) determining whether the voting value is greater than the minimum number of character;
b) increasing a silence timer if the voting value is greater than the minimum number of character;
c) determines whether the silence timer is greater than a threshold value of silence timer;
d) determining whether the number of filtered license plates in the subgroup is greater than a minimum number of filtered license plates in the sub-group if the silence timer is greater than the threshold value of silence timer; and
e) publishing the characters with the highest frequency as the final recognition result if the number of filtered license plates in the subgroup is greater than a minimum number of filtered license plates in the sub-group.
The method as claimed in claim 11 , wherein the plurality of filtered license plates is rejected if:
a) the voting value is less than or equals to the minimum number of character;
b) the silence timer is less than or equals to the threshold value of silence timer; and
c) the number of filtered license plates in the sub-group is less than or equals to the minimum number of filtered license plates result in the sub-group.
13. The method as claimed in claim 9, wherein the step includes rejecting the plurality of filtered license plates, if the delay timer is less than or equals to the threshold value of delay timer.
PCT/MY2018/050063 2017-09-29 2018-09-28 A system and method for recognising license plate Ceased WO2019066641A2 (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
MYPI2017001429 2017-09-29
MYPI2017001429 2017-09-29

Publications (2)

Publication Number Publication Date
WO2019066641A2 true WO2019066641A2 (en) 2019-04-04
WO2019066641A3 WO2019066641A3 (en) 2019-06-27

Family

ID=65861666

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/MY2018/050063 Ceased WO2019066641A2 (en) 2017-09-29 2018-09-28 A system and method for recognising license plate

Country Status (1)

Country Link
WO (1) WO2019066641A2 (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113239854A (en) * 2021-05-27 2021-08-10 北京环境特性研究所 Ship identity recognition method and system based on deep learning

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20130259314A1 (en) 2012-03-30 2013-10-03 Xerox Corporation Methods and systems for enhancing the performance of automated license plate recognition applications utilizing multiple results

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9785855B2 (en) * 2015-12-17 2017-10-10 Conduent Business Services, Llc Coarse-to-fine cascade adaptations for license plate recognition with convolutional neural networks

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20130259314A1 (en) 2012-03-30 2013-10-03 Xerox Corporation Methods and systems for enhancing the performance of automated license plate recognition applications utilizing multiple results

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113239854A (en) * 2021-05-27 2021-08-10 北京环境特性研究所 Ship identity recognition method and system based on deep learning
CN113239854B (en) * 2021-05-27 2023-12-19 北京环境特性研究所 Ship identity recognition method and system based on deep learning

Also Published As

Publication number Publication date
WO2019066641A3 (en) 2019-06-27

Similar Documents

Publication Publication Date Title
US20170026390A1 (en) Identifying Malware Communications with DGA Generated Domains by Discriminative Learning
MX2022003736A (en) Automatic license plate recognition (alpr) and vehicle identification profile methods and systems.
EP2747431B1 (en) Device and method for detecting whether camera is interfered with, and video monitoring system
CN103679018A (en) Method and device for detecting CSRF loophole
CN111898581A (en) Animal detection method, device, electronic equipment and readable storage medium
CN110287889A (en) A method and device for identification
CN109639744A (en) A kind of detection method and relevant device in the tunnel DNS
US20180268634A1 (en) Folded bill identification method and device
WO2011084130A4 (en) Human interaction trajectory-based system
KR101883228B1 (en) Method and Apparatus for Gesture Recognition
US20180173939A1 (en) Recognition of objects within a video
CN110414294A (en) Pedestrian recognition methods and device again
KR20200036079A (en) System and Method for Detecting Deep Learning based Human Object using Adaptive Thresholding Method of Non Maximum Suppression
CN110688883A (en) Vehicle and pedestrian detection method and device
CN115272847B (en) A target determination method and apparatus based on anomaly detection
CN113657315A (en) Method, device and equipment for screening quality of face image and storage medium
CN111126112A (en) Candidate region determination method and device
CN111027534A (en) A compact double license plate detection method and device
US20150150132A1 (en) Intrusion detection system false positive detection apparatus and method
WO2019066641A2 (en) A system and method for recognising license plate
KR101413620B1 (en) Apparatus for video to text using video analysis
CN112528774A (en) Intelligent sorting system and method for unknown radar signals in complex electromagnetic environment
US10394888B2 (en) Video search system and method
CN114283361B (en) Method and device for determining status information, storage medium and electronic device
KR102133330B1 (en) Apparatus and method for recognizing crack in sturcutre

Legal Events

Date Code Title Description
NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18859960

Country of ref document: EP

Kind code of ref document: A2