TWI494866B - License plate identification method and its hand - held electronic device - Google Patents

Description

License plate identification method and handheld electronic device thereof

The invention relates to a license plate identification method and a handheld electronic device thereof, in particular to a license plate recognition method capable of accurately recognizing the license plate image when a license plate image has a shadow, and a handheld electronic device thereof.

License plate recognition has been widely used in areas such as toll collection mechanisms for roads and parking lots, and road traffic monitoring and bans. The general license plate identification method mainly has the following steps. First, the image of the license plate is captured for an image, and then the characters on the license plate are sent one by one to a classifier (such as a neural network) for performing pre-image training to identify the license plate character recognition result.

In a sufficient light source environment, the correct rate of the license plate recognition method can reach a certain standard, but in an environment where the light is not uniformly illuminated, such as under the shade of the sunset at sunset, or in any environment obscured by the object in front, the license plate There will be shadows on it. In the binarization step of the license plate image affected by the shadow, the dark shadow area often fails to obtain the ideal binarization result. On the other hand, the license plate character in the unshaded area is compared with the license plate background color in the deep shadow area. The pixel value of the deep shadow area will be lower than the license plate character without the shadow area, so the result of the binarization is incorrect, and the result is large. The frame reduces the correct rate of license plate recognition.

Therefore, how to overcome the above problems makes the license plate identification method have a good accuracy rate in practical applications, which will be a major challenge for the current license plate identification method.

Accordingly, it is an object of the present invention to provide a license plate recognition method that provides accurate identification of a shaded license plate image.

Therefore, the license plate recognition method of the present invention is suitable for recognizing an image including a license plate image, and is implemented by a license plate recognition system, which includes a license plate image capturing unit, a shadow recognition unit, and a shadow filter. a unit, a license plate image processing unit, a character matching unit and a character database, the method comprising the following steps: (A) the license plate image capturing unit extracts the license plate image in the image; (B) The shadow recognition unit identifies, from the license plate image, a shadow recognition image including a shadow area and a non-shadow area, a non-shadow area image, and a shadow area image, wherein the non-shadow area image includes a license plate corresponding to the non-shadow area a non-shaded portion of the image, and the shadow region image includes a shadow portion of the license plate image corresponding to the shadow region; (C) the shadow filtering unit filters out shadow noise of the shadow portion of the license plate image of the shadow region image and outputs a shadow Filtering the image; (D) the license plate image processing unit generates a waiting image according to the unshaded area image and the shadow filtering image For the license plate image, wherein the to-be-matched license plate image has a plurality of character images; and (E) the character matching unit compares the images according to the plurality of characters in the character database, and compares the waiting ratios The character represented by each character image in the license plate image.

Another object of the present invention is to provide a license plate recognition handheld electronic device that provides accurate identification of a shaded license plate image.

Therefore, the license plate recognition handheld electronic device of the present invention is suitable for recognizing an image including a license plate image, wherein the license plate recognition handheld electronic device comprises a character database, a license plate image capturing unit, a shadow recognition unit, and a A shadow filtering unit, a license plate image processing unit, and a character matching unit. The character database is used to store multiple characters to compare images. The license plate image capturing unit is configured to extract the license plate image in the image. The shadow recognition unit is configured to identify, from the license plate image, a shadow recognition image including a shadow area and a non-shadow area, a non-shadow area image, and a shadow area image, wherein the non-shadow area image includes a corresponding one of the non-shadow areas The license plate image of the area is non-shaded, and the shadow area image includes a shaded portion of the license plate image corresponding to the shaded area. The shadow filtering unit is configured to filter out the shadow noise of the shadow portion of the license plate image of the shadow area image and output a shadow filtering image. The license plate image processing unit is configured to generate a to-be-matched license plate image according to the unshaded area image and the shadow filtered image, wherein the to-be-matched license plate image has a plurality of character images. The character matching unit is configured to compare the characters represented by each character image in the to-be-matched license plate image according to the characters in the character database.

The effect of the present invention is that the shadow recognition unit can recognize the non-shadow area image and the shadow area image, and filter the shadow noise of the shadow part of the license plate image of the shadow area image by the shadow filtering unit. The image of the to-be-matched license plate image has no shadow noise and thus improves the recognition rate.

1‧‧‧ License plate identification handheld electronic device

10‧‧‧ License Plate Identification System

11‧‧‧ License Plate Image Capture Unit

12‧‧‧Shadow recognition unit

13‧‧‧Shadow filter unit

14‧‧‧ License Plate Image Processing Unit

15‧‧‧ character matching unit

16‧‧‧ character database

21~27‧‧‧Steps

71~75‧‧‧Steps

800‧‧‧ License Plate Image

801‧‧‧Edge detection image

802‧‧‧ Shadow Recognition Image

803‧‧‧Unshaded area image

804‧‧‧ Shadow area image

805‧‧‧ Shadow filtering image

806‧‧‧Compare the license plate image

807‧‧‧Shaded area

808‧‧‧Unshaded area

809‧‧‧ Licensed image non-shaded

810‧‧‧ Shadow image of the license plate image

Other features and effects of the present invention will be apparent from the following description of the drawings. FIG. 1 is a system architecture diagram illustrating a preferred embodiment of the license plate recognition handheld electronic device of the present invention. An image containing a license plate image is identified; FIG. 2 is a flow chart illustrating a preferred embodiment of the license plate recognition method of the present invention for identifying an image containing the license plate image; FIG. 3 is a schematic diagram illustrating execution The main image generated by the license plate recognition method of the present invention; and FIG. 4 is a flowchart illustrating the application of the license plate recognition method of the present invention to the drive-through service.

Referring to FIG. 1 , a preferred embodiment of the license plate recognition handheld electronic device 1 of the present invention is suitable for identifying an image including a license plate image, such as a smart phone, a PDA, a tablet computer. The system is implemented by a license plate recognition system 10 in the form of a software placed therein. The license plate recognition system 10 includes a license plate image capturing unit 11, a shadow recognition unit 12, a shadow filtering unit 13, a license plate image processing unit 14, a character matching unit 15, and a character database 16.

The character database 16 stores a plurality of character alignment images.

The license plate image capturing unit 11 picks up the car in the image Card image. In the preferred embodiment, the license plate image capturing unit 11 is mainly implemented by an AdaBoost algorithm based on a license plate and a character training sample, but the license plate image capturing unit 11 can also be used. The implementation of the license plate location algorithm is not limited to the preferred embodiment. Since the detailed implementation of the license plate image capturing unit 11 is not the focus of the patent application, it will not be described here. For a detailed implementation of the license plate image capturing unit 11, reference may be made to "based on license plate and character training. Sample AdaBoost license plate detection technology and its Android license plate recognition driving recorder implementation" this paper.

The shadow recognition unit 12 performs binarization processing on the license plate image and performs edge detection to generate an edge detection image having a plurality of edge pixels, and performs shape expansion and filling operation on the edge detection image. And morphologically eroding to generate the shadow recognition image identifying the shadow area and the non-shadow area, and identifying a non-shadow area image according to the shadow recognition image, a non-shaded portion of a license plate image, and a shadow portion of a license plate image a shaded area image, wherein the unshaded area image includes a non-shaded portion of the license plate image corresponding to the unshaded area, and the shadow area image includes a shaded portion of the license plate image corresponding to the shaded area.

The shadow filtering unit 13 uses the fast homomorphic filter and the contrast enhancement processing to filter the shadow noise of the shadow portion of the license plate image of the shadow region image and output a shadow filtering image.

The license plate image processing unit 14 performs a joint operation on the unshaded area image and the shadow filtered image to generate a to-be-matched license plate image, wherein the to-be-matched license plate image has a plurality of character images.

The character matching unit 15 compares the images according to the characters in the character database 16 and compares the characters represented by each character image in the to-be-matched license plate image. In the preferred embodiment, the Tesseract OCR character recognition software is used to compare the license plate characters. The Tesseract OCR character recognition software is a set of open source software, due to its components, identification principles, and details thereof. The manner of implementation is well known to those skilled in the art and will not be described here.

Referring to FIG. 1, FIG. 2 and FIG. 3, a preferred embodiment of the license plate recognition method of the present invention is suitable for identifying an image (not shown) including the license plate image 800, and includes the following steps.

As shown in step 21, the license plate image capturing unit 11 extracts the license plate image 800 in the image.

As shown in step 22, the shadow recognition unit 12 binarizes the license plate image 800 and performs edge detection to generate the edge detection image 801 having the edge pixels.

As shown in step 23, the shadow recognition unit 12 performs a shape expansion, a filling operation, and a morphological erosion on the edge detection image 801 to generate the shadow recognition image 802 that identifies the shadow region 807 and the non-shadow region 808. .

As shown in step 24, the shadow recognition unit 12 identifies the non-shadow area image 803 and the shadow area image 804 according to the shadow recognition image 802, the license plate image non-shadow portion 809, and the license plate image shadow portion 810, wherein the non-shadow area image 803 includes a license plate image non-shaded portion 809 corresponding to the non-shaded area 808, and the shadow area image 804 package A license plate image shaded portion 810 corresponding to the shaded area 807 is included.

As shown in step 25, the shadow filtering unit 13 filters the shadow noise of the shaded image portion 810 of the shaded image 804 by the fast homomorphic filter and the contrast enhancement processing and outputs the shadow filtered image 805.

As shown in step 26, the license plate image processing unit 14 performs a union operation on the unshaded area image 803 and the shadow filtered image 805 to generate the to-be-matched license plate image 806, wherein the to-be-matched license plate image 806 has such Character image.

As shown in step 27, the character matching unit 15 compares the images according to the characters in the character database 16 and compares the image represented by each character image in the to-be-matched license plate image 806. Character.

Referring to FIG. 1 and FIG. 4, the license plate identification method of the present invention can be applied to various types of speed-of-service services (for example, in fast food restaurants, coffee shops, pharmacies, telecommunications company service windows, etc.) and various types of vehicles. In the application of license plate recognition. For example, the drive-through service used in fast food restaurants, the drive-free service is mainly to provide a dedicated lane to the consumers who drive, so that consumers do not need to find a parking space and do not have to get off. For ease of understanding, the following describes the ordering process of the consumer, the processing steps of the drive-through ordering system, and the license plate recognition step, but only the license plate recognition step belongs to the steps included in the method of the present invention.

First, as shown in step 71, the consumer pre-orders the merchandise in the pick-and-roll ordering system (not shown in FIG. 1) using its consumer electronic device (not shown in FIG. 1) and the license plate number to trigger the deferred order. System saves a subscription Recorded separately in the drive-through ordering system, wherein the order record includes the license plate number and the item ordered by the consumer.

Next, as shown in step 72, when the consumer drives his vehicle into the dedicated lane, the license plate recognition system 10 obtains the image containing the license plate image. The image may be obtained by capturing a image of a license plate of the vehicle by a photographing device and transmitting it to the license plate recognition system 10, or by identifying the handheld electronic device 1 by the license plate recognition system 10. The image is captured by the photographing function of the license plate of the vehicle.

Next, as shown in step 73, the license plate recognition system 10 performs the steps included in the license plate recognition method of the present invention to compare the characters represented by each character image in the license plate image.

Next, as shown in step 74, the license plate recognition function of the handheld electronic device 1 is carried out by the license plate on which the license plate recognition system 10 is placed, and the comparison result of the license plate image is transmitted to the drive-through ordering system.

Finally, as shown in step 75, the drive-through ordering system compares the order record of the license plate image with the order record stored in the drive-through ordering system, and compares the order record associated with the license plate number, and The order record is displayed on a screen located at a drive-through service provider (for example, a kitchen in a fast food restaurant).

Drive-through services are usually divided into three phases: (1) placing orders; (2) paying; and (3) picking up goods. However, the bottlenecks that cost the most time and lead to long-running consumers are all in the "order" stage. This is because consumers and service providers often need to spend a lot during the "order" stage. Time to communicate and confirm the content of the order. By applying the license plate recognition method of the present invention to the drive-through service of the fast food restaurant, so that the consumer can use the license plate number to pre-order the goods in the drive-through ordering system, thereby omitting the service in the drive-through service. The communication time required for a single phase.

In addition, the license plate electronic wallet can be combined, so that after the ordering of the goods is completed, the stored value in the license plate electronic wallet can be debited and settled, so that the consumer can further omit the cost of the paying phase in the driving service. Time to find money.

In summary, the license plate recognition method of the present invention recognizes the unshaded area image 803 and the shadow area image 804 by the shadow recognition unit 12, and filters the license plate image of the shadow area image 804 by the shadow filtering unit 13. The shadow noise of the shaded portion 810 makes the to-be-matched license plate image 806 have no shadow noise and thus improves the correct rate of license plate recognition, so the object of the present invention can be achieved.

The above is only the preferred embodiment of the present invention, and the scope of the present invention is not limited thereto, that is, the simple equivalent changes and modifications made by the patent application scope and patent specification content of the present invention, All remain within the scope of the invention patent.

21~27‧‧‧Steps

Claims (8)

A license plate recognition method is suitable for identifying an image including a license plate image, and is implemented by a license plate recognition system, which includes a license plate image capturing unit, a shadow recognition unit, and a shadow filtering unit. a license plate image processing unit, a character matching unit and a character database, the method comprising the following steps: (A) the license plate image capturing unit extracts the license plate image in the image; (B) the shadow recognition The unit identifies, from the license plate image, a shadow recognition image including a shadow area and a non-shadow area, a non-shadow area image, and a shadow area image, wherein the non-shadow area image includes a license plate image corresponding to the non-shadow area. a shaded portion, and the shadow region image includes a shadow portion of the license plate image corresponding to the shadow region; (C) the shadow filtering unit filters out shadow noise of the shadow portion of the license plate image of the shadow region image and outputs a shadow filter (D) the license plate image processing unit generates a waiting vehicle according to the unshaded area image and the shadow filtered image An image, wherein the to-be-matched license plate image has a plurality of character images; and (E) the character matching unit compares the plurality of characters in the character database to compare the images with the pair of license plates The character represented by each character image in the image. The license plate recognition method according to claim 1, wherein the step (B) comprises the following sub-steps: (B-1) the shadow recognition unit binarizes the license plate image and performs edge detection to generate an edge detection image having a plurality of edge pixels; (B-2) the shadow recognition unit The edge detection image performs a shape expansion, a filling operation, and a shape erosion to generate the shadow recognition image that identifies the shadow area and the non-shadow area; and (B-3) the shadow recognition unit identifies the image according to the shadow, The non-shaded portion of the license plate image and the shaded portion of the license plate image identify the unshaded area image and the shadow area image. The license plate recognition method according to claim 1, wherein in the step (C), the shadow filtering unit filters the shadow noise of the shadow region image by using a fast homomorphic filter and a contrast enhancement processing. The license plate recognition method according to claim 1, wherein in the step (D), the license plate image processing unit performs a joint operation on the unshaded area image and the shadow filtered image to generate the to-be-matched license plate image. A license plate recognition handheld electronic device is adapted to identify an image including a license plate image. The license plate recognition handheld electronic device comprises: a character database for storing a plurality of character comparison images; a license plate image a capture unit for extracting the license plate image in the image; a shadow recognition unit for recognizing a shadow recognition image including a shadow region and a non-shadow region, and a non-shadow region image from the license plate image And a shadow area image, wherein the non-shadow area image Included is a non-shaded portion of the license plate image corresponding to the unshaded area, and the shadow area image includes a shadow portion of the license plate image corresponding to the shadow area; and a shadow filtering unit for filtering the image of the license plate image of the shadow area image Part of the shadow noise and outputting a shadow filtering image; a license plate image processing unit for generating a to-be-matched license plate image according to the non-shadow area image and the shadow filtering image, wherein the to-be-matched license plate image has multiple a character image; and a character matching unit for comparing the words represented by each character image in the to-be-matched license plate image according to the characters in the character database yuan. The license plate recognition handheld electronic device of claim 5, wherein the shadow recognition unit is configured to perform binarization processing on the license plate image and perform edge detection to generate an edge detection image having a plurality of edge pixels. And performing the pattern expansion, filling operation, and shape erosion on the edge detection image to generate the shadow recognition image that identifies the shadow area and the non-shadow area, and identifying the image and the non-shaded part of the license plate image according to the shadow And the shaded portion of the license plate image identifies the unshaded area image and the shadow area image. The license plate recognition handheld electronic device according to claim 5, wherein the shadow filtering unit is configured to filter the shadow noise of the shadow region image by using a fast homomorphic filter and a contrast enhancement process. The license plate recognition handheld electronic device according to claim 5, wherein the license plate image processing unit is configured to use the shadow image and the shadow filter The image is subjected to a union operation to generate the to-be-matched license plate image.