License plate reading apparatus and method
First Claim
1. A license plate reading apparatus comprising:
- an image pickup unit, provided above a road, to photograph a running vehicle at a predetermined frame period;
an image storing unit to successively store two images of a present frame image and a previous frame image photographed by said image pickup unit;
a vehicle detecting unit to detect a motion of a license plate in a front portion of said running vehicle on the basis of a difference between the present frame image and the previous frame image stored in said image storing unit;
a dictionary in which character images of all characters used for the license plate have been registered; and
a character recognizing unit to recognize all of the characters on the license plate from an image in a license plate area detected by said vehicle detecting unit by collating with the character images registered in said character image dictionary, wherein said vehicle detecting unit comprises;
a block dividing unit to divide each of said present frame image and said previous frame image into blocks of a predetermined pixel size;
a vector calculating unit to calculate a motion vector of the block by obtaining a pixel position, as a block moving position, in which a sum of differences between each pixel in the block of said previous frame image and all pixels in the block of said present frame image becomes the minimum; and
a vehicle head detecting unit to scan all of the blocks, to search a group of blocks in which sizes and directions of the motion vectors are almost the same in an area of a rectangular size corresponding to the license plate, to detect the vehicle on the basis of the presence or absence of said group of blocks which satisfy said conditions, and at the same time, to detect a license plate area at the head of the vehicle.
2 Assignments
0 Petitions
Accused Products
Abstract
A running vehicle is photographed at a predetermined frame period by an image pickup unit (TV camera) provided at a position above a road. Two images of a present frame image and a previous frame image which were photographed are successively stored in an image storing unit. A vehicle detecting unit detects a motion of a license plate in a head portion of the running vehicle on the basis of the present frame image and the previous frame image stored in the image storing unit. A character recognizing unit recognizes all of the characters on the license plate on the basis of a dictionary in which character images of all of the characters used for the license plate have been registered and an image in a license plate area detected by the vehicle detecting unit by collating with the character images registered in the dictionary.
88 Citations
8 Claims
-
1. A license plate reading apparatus comprising:
-
an image pickup unit, provided above a road, to photograph a running vehicle at a predetermined frame period;
an image storing unit to successively store two images of a present frame image and a previous frame image photographed by said image pickup unit;
a vehicle detecting unit to detect a motion of a license plate in a front portion of said running vehicle on the basis of a difference between the present frame image and the previous frame image stored in said image storing unit;
a dictionary in which character images of all characters used for the license plate have been registered; and
a character recognizing unit to recognize all of the characters on the license plate from an image in a license plate area detected by said vehicle detecting unit by collating with the character images registered in said character image dictionary, wherein said vehicle detecting unit comprises;
a block dividing unit to divide each of said present frame image and said previous frame image into blocks of a predetermined pixel size;
a vector calculating unit to calculate a motion vector of the block by obtaining a pixel position, as a block moving position, in which a sum of differences between each pixel in the block of said previous frame image and all pixels in the block of said present frame image becomes the minimum; and
a vehicle head detecting unit to scan all of the blocks, to search a group of blocks in which sizes and directions of the motion vectors are almost the same in an area of a rectangular size corresponding to the license plate, to detect the vehicle on the basis of the presence or absence of said group of blocks which satisfy said conditions, and at the same time, to detect a license plate area at the head of the vehicle.
-
-
2. A license plate reading apparatus comprising:
-
an image pickup unit, provided above a road, to photograph a running vehicle at a predetermined frame period;
an image storing unit to successively store two images of a present frame image and a previous frame image photographed by said image pickup unit;
a vehicle detecting unit to detect a motion of a license plate in a front portion of said running vehicle on the basis of a difference between the present frame image and the previous frame image stored in said image storing unit;
a dictionary in which character images of all characters used for the license plate have been registered; and
a character recognizing unit to recognize all of the characters on the license plate from an image in a license plate area detected by said vehicle detecting unit by collating with the character images registered in said character image dictionary wherein said vehicle detecting unit comprises;
a block dividing unit to divide each of said present frame image and said previous frame image into blocks of a predetermined pixel size;
a luminance level threshold value setting unit to always measure a luminance in a predetermined area on the road in said frame image, to obtain an average value for a predetermined time to thereby discriminate a statistical external illuminance state, to obtain a luminance level threshold value which increases monotonously in accordance with said illuminance state by a predetermined simple increasing curve, and to set said luminance level threshold value;
a vector calculating unit to calculate a motion vector of the block by obtaining a pixel position, as a block moving position, in which a sum of differences between each pixel in the block of said previous frame image and all pixels in the block of said present frame image becomes the minimum;
a reliability amount calculating unit to calculate a motion vector reliability amount by subtracting the sum of the differences between each pixel in the block of the present frame image and all pixels in the block of said previous frame image from a minimum difference value calculated by said vector calculating unit; and
a vehicle head detecting unit to scan all of the blocks, to search a group of blocks in which sizes and directions of the motion vectors are almost the same and said reliability amount is equal to or larger than said luminance level threshold value in an area of a rectangular size corresponding to the license plate, to detect the vehicle on the basis of the presence or absence of said group of blocks which satisfy said conditions, and at the same time, to detect a license plate area at the head of the vehicle.
-
-
3. A license plate reading apparatus comprising:
-
an image pickup unit, provided above a road, to photograph a running vehicle at a predetermined frame period;
an image storing unit to successively store two images of a present frame image and a previous frame image photographed by said image pickup unit;
a vehicle detecting unit to detect a motion of a license plate in a front portion of said running vehicle on the basis of a difference between the present frame image and the previous frame image stored in said image storing unit;
a dictionary in which character images of all characters used for the license plate have been registered; and
a character recognizing unit to recognize all of the characters on the license plate from an image in a license plate area detected by said vehicle detecting unit by collating with the character images registered in said character image dictionary;
wherein said character recognizing unit comprises;
a low resolution recognizing unit to recognize a character at a position of the character having the largest size on the license plate with respect to a frame image of a low resolution, as a target, obtained by thinning out an original image; and
a high resolution recognizing unit to restore said frame image to an original resolution, to recognize the character at the character position recognized by said low resolution recognizing unit, and to recognize another character at a position of the character having a small size on the license plate where a relative position has previously been known on the basis of a result of said character recognition. - View Dependent Claims (4)
in said dictionary, a sample character image obtained by photographing each character which is used for the license plate, is registered, and each of said sample character images is reduced to an image corresponding to a size of character image at the time of reading the license plate by thinning out said image to an image of a same resolution as that of said low resolution recognizing unit and, thereafter, converted into a Laplacian character image by applying a Laplacian filter, and registered, and each of said low resolution recognizing unit and said high resolution recognizing unit converts an inputted character image into the image by applying the Laplacian filter and, thereafter, performs a normalization correlation arithmetic operation between said converted image and said Laplacian character image of the corresponding resolution which has been registered in the dictionary, thereby recognizing the characters of the license plate.
-
-
5. A license plate reading method comprising:
-
a dictionary forming step of forming a dictionary in which character images of all characters used for a license plate have been registered;
an image pickup step of photographing a running vehicle at a predetermined frame period from a position above a road;
an image storing step of successively storing two images of a present frame image and a previous frame image photographed in said image pickup step;
a vehicle detecting step of detecting a motion of the license plate in a front portion of said running vehicle on the basis of a difference between the present frame image and the previous frame image stored by said image storing step; and
a character recognizing step of recognizing all of the characters on the license plate from an image in a license plate area detected in said vehicle detecting step by collating with the character images registered in said dictionary;
wherein said vehicle detecting step comprises;
a block dividing step of dividing each of said present frame image and said previous frame image into blocks of a predetermined pixel size;
a vector calculating step of calculating a motion vector of the block by obtaining a pixel position, as a block moving position, in which a sum of differences between each pixel in the block of said previous frame image and all pixels in the block of said present frame image becomes the minimum; and
a vehicle head detecting step of scanning all of the blocks, searching a group of blocks in which sizes and directions of the motion vectors are almost the same in an area of a rectangular size corresponding to the license plate, detecting the vehicle on the basis of the presence or absence of said group of blocks which satisfy said conditions, and at the same time, detecting a license plate area at the head of the vehicle.
-
-
6. A license plate reading method comprising:
-
a dictionary forming step of forming a dictionary in which character images of all characters used for a license plate have been registered;
an image pickup step of photographing a running vehicle at a predetermined frame period from a position above a road;
an image storing step of successively storing two images of a present frame image and a previous frame image photographed in said image pickup step;
a vehicle detecting step of detecting a motion of the license plate in a front portion of said running vehicle on the basis of a difference between the present frame image and the previous frame image stored by said image storing step; and
a character recognizing step of recognizing all of the characters on the license plate from an image in a license plate area detected in said vehicle detecting step by collating with the character images registered in said dictionary;
wherein said vehicle detecting step comprises;
a block dividing step of dividing each of said present frame image and said previous frame image into blocks of a predetermined pixel size;
a luminance level threshold value setting step of always measuring a luminance in a predetermined area on the road in said frame image, obtaining an average value for a predetermined time to thereby discriminate a statistical external illuminance state, obtaining a luminance level threshold value which increases monotonously in accordance with said illuminance state by a predetermined simple increasing curve, and setting said luminance level threshold value;
a vector calculating step of calculating a motion vector of the block by obtaining a pixel position, as a block moving position, in which a sum of differences between each pixel in the block of said previous frame image and all pixels in the block of said present frame image becomes the minimum;
a reliability amount calculating step of calculating a motion vector reliability amount by subtracting the sum of the differences between each pixel in the block of the present frame image and all pixels in the block of said previous frame image from a minimum difference value calculated by said vector calculating unit; and
a vehicle head detecting step of scanning all of the blocks, searching a group of blocks in which sizes and directions of the motion vectors are almost the same and said reliability amount is equal to or larger than said luminance level threshold value in an area of a rectangular size corresponding to the license plate, detecting the vehicle on the basis of the presence or absence of said group of blocks which satisfy said conditions, and at the same time, detecting a license plate area at the head of the vehicle.
-
-
7. A license plate reading method comprising:
-
a dictionary forming step of forming a dictionary in which character images of all characters used for a license plate have been registered;
an image pickup step of photographing a running vehicle at a predetermined frame period from a position above a road;
an image storing step of successively storing two images of a present frame image and a previous frame image photographed in said image pickup step;
a vehicle detecting step of detecting a motion of the license plate in a front portion of said running vehicle on the basis of a difference between the present frame image and the previous frame image stored by said image storing step; and
a character recognizing step of recognizing all of the characters on the license plate from an image in a license plate area detected in said vehicle detecting step by collating with the character images registered in said dictionary;
wherein said character recognizing step comprises;
a low resolution recognizing step of recognizing a character at a position of the character having the largest size on the license plate with respect to a frame image of a low resolution, as a target, obtained by thinning out an original image; and
a high resolution recognizing step of restoring said frame image to an original resolution, recognizing the character at the character position recognized in said low resolution recognizing step, and recognizing another character at a position of the character having a small size on the license plate where a relative position has previously been known on the basis of a result of said character recognition. - View Dependent Claims (8)
in said dictionary forming step, a sample character image obtained by photographing each character which is used for the license plate, is registered in said dictionary, and each of said sample character images is reduced to an image corresponding to a size of character image at the time of reading the license plate by thinning out said image to an image of a same resolution as that in said low resolution recognizing step and, thereafter, converted into a Laplacian character image by applying a Laplacian filter, and registered in said dictionary, and in each of said low resolution recognizing step and said high resolution recognizing step, an inputted character image is converted into the image by applying the Laplacian filter and, thereafter, a normalization correlation arithmetic operation is performed between said converted image and said Laplacian character image of the corresponding resolution which has been registered in said dictionary, thereby recognizing the characters of the license plate.
-
Specification