WO2014076979A1 - 画像認識装置、画像認識方法、プログラム、及び記録媒体 - Google Patents
画像認識装置、画像認識方法、プログラム、及び記録媒体 Download PDFInfo
- Publication number
- WO2014076979A1 WO2014076979A1 PCT/JP2013/059061 JP2013059061W WO2014076979A1 WO 2014076979 A1 WO2014076979 A1 WO 2014076979A1 JP 2013059061 W JP2013059061 W JP 2013059061W WO 2014076979 A1 WO2014076979 A1 WO 2014076979A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- number plate
- edge
- distribution
- image
- vehicle number
- 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
Links
Images
Classifications
-
- 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
-
- 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
- G06V20/63—Scene text, e.g. street names
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- 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 an image recognition device, an image recognition method, a program, and a recording medium.
- the present invention relates to an image recognition device that extracts a region of a car number plate from an image, an image recognition method, a program that causes a computer to function as the image recognition device, and a recording medium that records the program.
- the car number information is a key for identifying individual vehicles, it has been applied to various purposes.
- the vehicle number is a character string assigned to each vehicle.
- Patent Documents 1 and 2 Various techniques related to such a background are known (see, for example, Patent Documents 1 and 2).
- Patent Document 1 describes a device that automatically identifies the vehicle number of a vehicle that is traveling on a road or is stopped. More specifically, this apparatus photographs a road. And this apparatus determines the candidate of a vehicle number plate frame from the image
- the vehicle number plate position can be determined from the degree of overlap between the vehicle number plate frame candidate and the vehicle number region candidate, so the determination of the vehicle number plate position can be made only from any one of the candidates.
- the vehicle number plate cutting process can be performed with a more certain probability.
- Patent Document 2 describes a device that performs vehicle number plate recognition. More specifically, this apparatus reduces the captured image in the horizontal direction and the vertical direction. Then, this apparatus subtracts the value of each pixel in the reduced image from the value of each pixel by a predetermined amount shifted in the positive or negative direction from the pixel, and the subtraction result is used as a binarization threshold value. A binarized comparison is made, and a pseudo shifted correlation image is generated by a logical product operation of the value shifted in the positive direction and the value shifted in the negative direction.
- this apparatus subtracts the value of each pixel in the reduced image from the value of each pixel by a predetermined amount shifted in the positive or negative direction from the pixel, and the subtraction result is used as a binarization threshold value.
- the binarized values are compared, and a pseudo-OR image is generated by a logical OR operation of the value shifted in the positive direction and the value shifted in the negative direction.
- this apparatus divides the pseudo-shifted correlation image into a plurality of small areas, calculates the area of a pixel whose luminance value is set to 1 in each small area, and based on the area value of each small area Select the car number plate candidate small area.
- this device temporarily sets a vehicle number plate region on the pseudo-OR image based on the vehicle number plate candidate small region, and checks the suitability as the vehicle number plate region for this temporary setting region. Cut out the number plate area.
- the vehicle number plate region can be cut out at high speed and with high accuracy using a general-purpose arithmetic device without newly developing a dedicated arithmetic device, and the development period can be shortened, and Cost can be reduced.
- the target vehicles are increasing. Therefore, in the vehicle number recognition system, in order to recognize the vehicle number at a higher speed, it is required to extract the region of the vehicle number plate from which characters are to be extracted at a higher speed.
- an image recognition device for extracting a region of a vehicle number plate from an image, the direction of the outline of a character that can be used for the vehicle number plate
- An edge direction evaluation unit that assigns points to the coordinates and evaluates the direction of the edge at the coordinates, and an area of the car number plate is extracted from the image based on the distribution of the points provided by the edge direction evaluation unit
- a vehicle number plate area extracting unit A vehicle number plate area extracting unit.
- the image processing apparatus further includes an edge strength determination unit that determines whether or not the edge strength at each coordinate of the image is larger than a predetermined strength threshold, and the edge direction evaluation unit has the edge strength of the predetermined strength. If the edge strength determination unit determines that the value is smaller than the threshold value, the coordinates are not affected even if the deviation amount between the contour direction and the edge direction is smaller than the predetermined deviation amount threshold value. It is not necessary to give a score to.
- the horizontal region of the vehicle number plate may be extracted from the image.
- a vertical direction distribution totaling unit that totals the vertical distribution of the points given by the edge direction evaluation unit is further provided, and the vehicle number plate area extraction unit is based on the distribution of the vertical points totalized by the vertical direction distribution totaling unit.
- a region in the vertical direction of the vehicle number plate may be extracted from the image.
- An area distribution totaling unit that totals the distribution of the points given by the edge direction evaluation unit for each predetermined region is further provided, and the vehicle number plate region extraction unit is based on the distribution of the points for each predetermined region totaled by the region distribution totaling unit.
- the area of the car number plate may be extracted from the image.
- the edge direction evaluation unit may give different points according to the existence ratio of the contour direction when the amount of deviation between the direction of the contour line and the direction of the edge is smaller than a predetermined deviation amount threshold value.
- a vehicle number plate region extraction method for extracting a region of a vehicle number plate from an image, wherein a predetermined presence is present among the directions of outlines of characters that can be used for the vehicle number plate.
- An edge direction evaluation stage for assigning points and evaluating the direction of the edge at the coordinates, and a vehicle number plate for extracting a region of the vehicle number plate from the image based on the distribution of the points assigned in the edge direction evaluation stage A region extraction stage.
- the deviation amount between the direction of the contour line having a higher presence ratio than the predetermined presence ratio threshold value and the edge direction at each coordinate of the image is smaller than the predetermined deviation amount threshold value.
- an edge direction evaluation unit that assigns points to the coordinates and evaluates the direction of the edge at the coordinates, and based on the distribution of the points provided by the edge direction evaluation unit, the region of the car number plate is extracted from the image. It functions as a vehicle number plate area extraction unit for extraction.
- a recording medium recording a program for causing a computer to function as an image recognition device for extracting a region of a vehicle number plate from an image, wherein the computer is used as a vehicle number plate.
- the amount of deviation between the direction of the outline of the outline of the character to be obtained that is higher than the threshold of the predetermined abundance and the direction of the edge at each coordinate of the image is a threshold of the predetermined deviation. If the value is smaller than the value, an edge direction evaluation unit that assigns a score to the coordinates and evaluates the direction of the edge at the coordinates, and a vehicle from the image based on the distribution of the points provided by the edge direction evaluation unit.
- a program that functions as a vehicle number plate area extracting unit that extracts a number plate area is recorded.
- the region of the vehicle number plate can be extracted from the image with high speed and accuracy.
- FIG. 1 shows an example of a use environment of a vehicle number recognition system 100 according to an embodiment.
- the vehicle number recognition system 100 is a system that takes a picture of the front part of a vehicle and recognizes characters on the vehicle number plate from the image.
- the vehicle number recognition system 100 includes an image recognition device 110, a CCTV (Closed Circuit TeleVision) camera 130, an auxiliary light source 150, and a host computer 170.
- image recognition device 110 a CCTV (Closed Circuit TeleVision) camera 130
- auxiliary light source 150 a host computer 170.
- the image recognition device 110 is a device that extracts the area of the vehicle number plate from the image.
- the image recognition device 110 is electrically connected to the CCTV camera 130. Further, the image recognition apparatus 110 is communicatively connected to the host computer 170 via the communication line N.
- the communication line N includes a computer network such as the Internet, a core network of a communication carrier, and various local networks.
- CCTV camera 130 is a device used for car number recognition.
- the CCTV camera 130 is electrically connected to the image recognition device 110.
- the CCTV camera 130 has sufficient resolution, sensitivity, and SN (Signal Noise) ratio for computer processing of images.
- the auxiliary light source 150 is a device that irradiates the driver with near-infrared illumination that is not illusion so that the vehicle number can be recognized even at night.
- the auxiliary light source 150 is electrically connected to the CCTV camera 130.
- the auxiliary light source 150 includes an LED (Light Emitting Diode), a xenon lamp, a halogen lamp, and the like.
- the host computer 170 is a device that performs processing using vehicle number information.
- the host computer 170 is communicatively connected to the image recognition apparatus 110 via a communication line N. Then, the host computer 170 uses the vehicle number information as a key to fix each vehicle, measures the time required to pass a specific section on the main road, determines the toll type on the toll road, and the contract vehicle in the parking lot Judgment, arrival monitoring of a specific vehicle, acquisition of usage history, acquisition of customer information, and quick discovery of stolen vehicles in emergency deployment.
- the vehicle number recognition system 100 includes a single image recognition device 110, CCTV camera 130, auxiliary light source 150, and host computer 170 for the purpose of preventing the explanation from becoming complicated. Will be described.
- the vehicle number recognition system 100 may include a plurality of image recognition devices 110, a CCTV camera 130, an auxiliary light source 150, and a host computer 170.
- FIG. 2 shows an example of a block configuration of the image recognition apparatus 110 according to the first embodiment.
- the image recognition device 110 includes an image input reception unit 111, a gradient calculation unit 112, an edge direction evaluation unit 113, a horizontal direction distribution tabulation unit 114, a vertical direction distribution tabulation unit 115, a car number plate region extraction unit 116, and a character recognition processing unit 117. And a car number data transmission unit 118.
- the function and operation of each component will be described in detail.
- the image input receiving unit 111 receives an image input output from the CCTV camera 130.
- the gradient calculation unit 112 calculates gradient values for the horizontal direction and the vertical direction of each coordinate of the image.
- calculating the value of the gradient can be said to be edge detection.
- An edge is a point where the pixel value changes sharply.
- the edge direction evaluation unit 113 includes a contour direction having a higher presence rate than a predetermined presence rate threshold value among the directions of character contour lines that can be used for the vehicle number plate, and the edge direction at each coordinate of the image.
- a predetermined presence rate threshold value among the directions of character contour lines that can be used for the vehicle number plate, and the edge direction at each coordinate of the image.
- the horizontal direction distribution totaling unit 114 totals the horizontal distribution of the points given by the edge direction evaluation unit 113.
- the vertical direction distribution totaling unit 115 totals the vertical distribution of the points given by the edge direction evaluation unit 113.
- the vehicle number plate region extraction unit 116 extracts the region of the vehicle number plate from the image based on the distribution of points given by the edge direction evaluation unit 113. For example, the vehicle number plate region extraction unit 116 extracts a horizontal region of the vehicle number plate from the image based on the distribution of the number of points in the horizontal direction counted by the horizontal direction distribution totalization unit 114. Further, for example, the vehicle number plate region extraction unit 116 extracts the vertical region of the vehicle number plate from the image based on the distribution of the score in the vertical direction totaled by the vertical direction distribution totalization unit 115.
- the character recognition processing unit 117 performs character recognition processing on the image of the vehicle number plate area to recognize the vehicle number.
- the vehicle number data transmission unit 118 transmits data indicating the vehicle number to the host computer 170.
- FIG. 3 shows an example of the distribution of the horizontal score tabulated by the horizontal distribution tabulation unit 114 in a graph format.
- the coordinate value in the horizontal direction is represented by i with respect to the image.
- the total value Vi_max is the maximum value of the horizontal score distribution.
- the horizontal coordinate i_max is the horizontal coordinate of the image that becomes the total value Vi_max.
- the total value Vi_threshold is a value obtained by multiplying the total value Vi_max by a predetermined coefficient ⁇ (provided that 0 ⁇ ⁇ 1).
- the horizontal coordinate i_left is a horizontal coordinate of the image in which the total value becomes the total value Vi_threshold for the first time when the distribution of the number of points in the horizontal direction is searched from the left end coordinate to the right end direction of the image.
- the horizontal coordinate i_right is the horizontal coordinate of the image in which the total value becomes the total value Vi_threshold for the first time when the distribution of the number of points in the horizontal direction is searched from the right end coordinate to the left end direction of the image.
- FIG. 4 shows an example of the distribution of the number of points in the vertical direction counted by the vertical direction distribution totaling unit 115 in a graph format.
- the coordinate value in the vertical direction is represented by j with respect to the image.
- the total value Vj_max is the maximum value of the score distribution in the vertical direction.
- the vertical coordinate j_max is the vertical coordinate of the image to be the total value Vj_max.
- the total value Vj_threshold is a value obtained by multiplying the total value Vj_max by a predetermined coefficient ⁇ (where 0 ⁇ ⁇ 1).
- the vertical coordinate j_upper is the vertical coordinate of the image in which the total value becomes the total value Vj_threshold for the first time when the distribution of the score in the vertical direction is searched from the upper end coordinate to the lower end direction of the image.
- the vertical coordinate j_lower is the vertical coordinate of the image in which the total value becomes the total value Vj_threshold for the first time when the distribution of the score in the vertical direction is searched from the lower end coordinate of the image to the upper end direction.
- FIG. 5 shows an example of the region of the vehicle number plate extracted by the vehicle number plate region extraction unit 116.
- the region of the vehicle number plate extracted by the vehicle number plate region extraction unit 116 is a region surrounded by start point coordinates (i_left, j_upper) and end point coordinates (i_right, j_lower).
- FIG. 6 shows an example of the operation flow of the image recognition apparatus 110. In the description of this operation flow, both FIGS. 1 to 5 are referred to.
- the CCTV camera 130 images a predetermined area on the road with a shutter speed of about 1/1000 second, for example.
- the auxiliary light source 150 emits light in synchronization with the shutter speed of the CCTV camera 130.
- CCTV camera 130 outputs the image obtained by imaging to the image recognition apparatus 110, whenever it images.
- the image input reception unit 111 of the image recognition apparatus 110 sends the image to the gradient calculation unit 112 and the character recognition processing unit 117 each time an input of an image output from the CCTV camera 130 is received (S101).
- the gradient calculation unit 112 of the image recognition apparatus 110 receives the image sent from the image input reception unit 111, the gradient value is obtained for each direction in the horizontal direction and the vertical direction of each coordinate while performing raster scanning of the image. Is calculated (S102). For example, the gradient calculation unit 112 calculates the horizontal gradient value f (row) of each coordinate of the image as in Expression (1). Further, for example, the gradient calculating unit 112 calculates the vertical gradient value f (col) of each coordinate of the image as in Expression (2).
- f (i, j) is a pixel value at coordinates (i, j).
- the gradient calculation unit 112 calculates the gradient value for each horizontal direction and vertical direction of each coordinate, the coordinate and the horizontal gradient value f (row) of the coordinate, Data indicating the gradient value f (col) in the vertical direction of the coordinates is sent to the edge direction evaluation unit 113.
- the edge direction evaluation unit 113 of the image recognition apparatus 110 receives the data sent from the gradient calculation unit 112, the edge direction evaluation unit 113 evaluates the edge direction at the coordinates indicated by the data (S103). For example, the edge direction evaluation unit 113 is based on the horizontal gradient value f (row) of the coordinate indicated by the data received from the gradient calculation unit 112 and the vertical gradient value f (col) of the coordinate. Then, the edge direction ⁇ at the coordinates is calculated as shown in Equation (3).
- the edge direction evaluation unit 113 determines whether the deviation amount between the edge direction ⁇ in the calculated coordinates and 0 ° or 90 ° is smaller than a predetermined deviation amount threshold value.
- the threshold value of the predetermined deviation amount is taken into account is that there is a possibility that the vehicle number plate is not always shown in the image horizontally or at an assumed angle.
- the threshold value of the predetermined deviation amount is, for example, about ⁇ 5 °.
- the edge direction evaluation unit 113 assigns one point to the coordinates. By giving, the direction of the edge at the coordinates is evaluated.
- the edge direction evaluation unit 113 does not assign points to the coordinates. Thus, the direction of the edge at the coordinates is evaluated.
- the edge direction evaluation unit 113 sends data indicating the coordinates and the number of points at the coordinates to the horizontal direction distribution totaling unit 114 and the vertical direction distribution totaling unit 115.
- the direction of the 0 ° outline and the direction of the 90 ° outline are defined as “a predetermined presence rate threshold value among the directions of the outlines of characters that can be used for the car number plate in the present invention. It may be an example of “the direction of the contour line with a high presence rate”.
- the horizontal direction distribution totaling unit 114 of the image recognition device 110 Upon receiving the data sent from the edge direction evaluation unit 113, the horizontal direction distribution totaling unit 114 of the image recognition device 110 totals the horizontal distribution of the points indicated by the data (S104). For example, the horizontal direction distribution totaling unit 114 adds the points of coordinates (i, j), coordinates (i, j + 1), coordinates (i, j + 2),. Value. In this way, when the horizontal direction distribution totalization unit 114 calculates the total value at each of the coordinates i, i + 1, i + 2,... In the horizontal direction, the data indicating the distribution of the points in the horizontal direction is obtained as the vehicle number plate. The data is sent to the area extraction unit 116.
- the vertical direction distribution totaling unit 115 of the image recognition device 110 Upon receiving the data sent from the edge direction evaluation unit 113, the vertical direction distribution totaling unit 115 of the image recognition device 110 totals the vertical distribution of the points indicated by the data (S105). For example, the vertical direction distribution totalization unit 115 adds the points of coordinates (i, j), coordinates (i + 1, j), coordinates (i + 2, j),. Value. In this way, when the vertical direction distribution totalization unit 115 calculates the total value at each of the coordinates j, j + 1, j + 2,... In the vertical direction, the data indicating the distribution of the score in the vertical direction is obtained as the vehicle number plate. The data is sent to the area extraction unit 116.
- the vehicle number plate region extraction unit 116 of the image recognition device 110 receives the data sent from the horizontal direction distribution totaling unit 114 and the vertical direction distribution totaling unit 115, respectively, based on the distribution of points indicated by these data.
- the area of the car number plate is extracted from the image (S106).
- the vehicle number plate region extracting unit 116 specifies the total value Vi_max that maximizes the horizontal score distribution from the horizontal score distribution indicated by the data received from the horizontal distribution totaling unit 114. .
- the vehicle number plate region extraction unit 116 calculates a value Vi_threshold obtained by multiplying the total value Vi_max by a predetermined coefficient ⁇ (where 0 ⁇ ⁇ 1).
- the car number plate region extraction unit 116 searches the distribution of the number of points in the horizontal direction from the left end coordinates of the image to the right end direction, and as shown in FIG. 3, the aggregate value is the first value Vi_threshold of the image.
- the horizontal coordinate i_left is specified.
- the car number plate region extraction unit 116 searches the distribution of the number of points in the horizontal direction from the coordinates of the right end of the image to the left end direction, and as shown in FIG. 3, the aggregate value is the first value Vi_threshold.
- the horizontal coordinate i_right is specified.
- the vehicle number plate region extraction unit 116 specifies the total value Vj_max that maximizes the distribution of the vertical score from the distribution of the vertical score indicated by the data received from the vertical distribution calculation unit 115. To do. Then, the vehicle number plate region extraction unit 116 calculates a value Vj_threshold obtained by multiplying the total value Vj_max by a predetermined coefficient ⁇ (where 0 ⁇ ⁇ 1). Then, the car number plate area extraction unit 116 searches the distribution of the score in the vertical direction from the coordinates of the upper end of the image to the direction of the lower end, and as shown in FIG. 4, as shown in FIG. The vertical coordinate j_upper is specified.
- the car number plate region extraction unit 116 searches the distribution of the score in the vertical direction from the coordinates of the lower end of the image toward the upper end, and as shown in FIG. The vertical coordinate j_lower is specified. Then, as shown in FIG. 5, the vehicle number plate region extraction unit 116 extracts a region surrounded by the start point coordinates (i_left, j_upper) and the end point coordinates (i_right, j_lower) as the region of the vehicle number plate. Then, the vehicle number plate region extraction unit 116 sends data indicating the start point coordinates (i_left, j_upper) and end point coordinates (i_right, j_lower) of the extracted vehicle number plate region to the character recognition processing unit 117.
- the character recognition processing unit 117 of the image recognition apparatus 110 receives the image sent from the image input receiving unit 111 and receives the data sent from the vehicle number plate area extracting unit 116
- the image receiving unit 111 receives the image received from the image input receiving unit 111.
- Character recognition processing for the area of the car number plate surrounded by the start point coordinates (i_left, j_upper) and the end point coordinates (i_right, j_lower) indicated by the data received from the car number plate area extracting unit 116 To recognize the vehicle number (S107).
- a pattern matching method capable of high-speed computation a method of extracting the feature amount of each character and separating it by distance calculation with a standard pattern, a method of directly giving a character pattern to a neural network, etc. are adopted. Is done. Then, the character recognition processing unit 117 sends data indicating the recognized vehicle number to the vehicle number data transmission unit 118.
- the vehicle number data transmission unit 118 of the image recognition device 110 transmits the data to the host computer 170 (S108).
- the host computer 170 can perform processing using the vehicle number information indicated by the data transmitted from the image recognition device 110.
- the image recognition device 110 has a presence rate contour that is higher than a predetermined presence rate threshold in the direction of the contour line of characters that can be used for the vehicle number plate.
- a predetermined presence rate threshold in the direction of the contour line of characters that can be used for the vehicle number plate.
- the image recognition device 110 can extract the region of the vehicle number plate from the image more quickly and accurately than the known technology.
- the image recognition apparatus 110 totals the horizontal distribution of the assigned points. And the image recognition apparatus 110 extracts the area
- the image recognition device 110 can extract the horizontal region of the vehicle number plate from the image with high speed and accuracy.
- the image recognition apparatus 110 totals the distribution of the assigned points in the vertical direction. And the image recognition apparatus 110 extracts the area
- the image recognition device 110 can extract the vertical region of the vehicle number plate from the image with high speed and accuracy.
- FIG. 7 shows an example of a block configuration of the image recognition apparatus 110 according to the second embodiment.
- the image recognition apparatus 110 according to the second embodiment includes an image input reception unit 111, a gradient calculation unit 112, an edge direction evaluation unit 113, a horizontal direction distribution tabulation unit 114, a vertical direction distribution tabulation unit 115, and a car number plate region extraction unit. 116, a character recognition processing unit 117, a vehicle number data transmission unit 118, and a lookup table 119.
- the function and operation of each component will be described in detail.
- constituent elements of the image recognition apparatus 110 according to the second embodiment the constituent elements having the same names as those of the constituent elements of the image recognition apparatus 110 according to the first embodiment are the same. Functions and operations are shown. Therefore, in the following description, the detailed description is abbreviate
- the look-up table 119 is an evaluation result of the edge direction ⁇ determined by each combination with respect to all combinations of the gradient value f (row) in the horizontal direction and the gradient value f (col) in the vertical direction. It is a table storing points. For example, the deviation amount between the edge direction ⁇ determined by the combination of the horizontal gradient value f (row) and the vertical gradient value f (col) and 0 ° or 90 ° is a predetermined deviation amount. When the value is smaller than the threshold value, one point is stored for the combination of the horizontal gradient value f (row) and the vertical gradient value f (col).
- the deviation amount between the edge direction ⁇ determined by the combination of the horizontal gradient value f (row) and the vertical gradient value f (col) and 0 ° or 90 ° is a predetermined deviation. If it is larger than the quantity threshold value, a score of 0 is stored for the combination of the horizontal gradient value f (row) and the vertical gradient value f (col).
- FIG. 8 shows an example of an operation flow of the image recognition apparatus 110 according to the second embodiment.
- both FIGS. 1 to 7 are referred to.
- the operation steps having the same names as those of the operation steps of the image recognition apparatus 110 according to the first embodiment have the same operations. Indicates. Therefore, in the following description, the detailed description is abbreviate
- the edge direction evaluation unit 113 of the image recognition apparatus 110 When the edge direction evaluation unit 113 of the image recognition apparatus 110 receives the data sent from the gradient calculation unit 112, the edge direction evaluation unit 113 indicates the data received from the gradient calculation unit 112 among the information on the number of points stored in the lookup table 119. The number of points stored for the combination of the horizontal gradient value f (row) and the vertical gradient value f (col) is read out. Then, the edge direction evaluation unit 113 uses the score read from the lookup table 119 as the evaluation result of the edge direction at the coordinates indicated by the data received from the gradient calculation unit 112 (S109). Then, the edge direction evaluation unit 113 sends data indicating the coordinates and the number of points at the coordinates to the horizontal direction distribution totaling unit 114 and the vertical direction distribution totaling unit 115.
- the image recognition apparatus 110 evaluates the edge direction with reference to the lookup table 119.
- the image recognition device 110 can extract the region of the vehicle number plate from the image at a higher speed.
- FIG. 9 shows an example of an operation flow of the image recognition apparatus 110 according to the third embodiment.
- both FIGS. 1 to 8 are referred to.
- the operation steps of the image recognition apparatus 110 according to the third embodiment the operation steps having the same names as those of the operation steps of the image recognition apparatus 110 according to the first embodiment have the same operations. Indicates. Therefore, in the following description, the detailed description is abbreviate
- the edge direction evaluation unit 113 of the image recognition apparatus 110 receives the data sent from the gradient calculation unit 112
- the edge direction evaluation unit 113 evaluates the edge direction at the coordinates indicated by the data (S110). For example, the edge direction evaluation unit 113, based on the horizontal gradient value f (row) indicated by the data received from the gradient calculation unit 112 and the vertical gradient value f (col) of the coordinates, The edge direction ⁇ at the coordinates is calculated as shown in Equation (3).
- characters that can be used in the vehicle number plate are represented in a predetermined font. Therefore, the direction of the outline of the character that can be used for the vehicle number plate tends to be biased in a specific direction.
- the direction of the contour line of 0 ° and the direction of the contour line of 90 ° have a particularly high existence rate.
- the direction of the 45 ° contour line is assumed to be the second highest presence rate next to the direction of the 0 ° or 90 ° contour line.
- the edge direction evaluation unit 113 determines whether the deviation amount between the edge direction ⁇ in the calculated coordinates and 0 °, 45 °, or 90 ° is smaller than a predetermined deviation amount threshold value. Determine whether.
- the reason why the threshold value of the predetermined deviation amount is taken into account is that there is a possibility that the vehicle number plate is not always shown horizontally in the image.
- the threshold value of the predetermined deviation amount is, for example, about ⁇ 5 °.
- the edge direction evaluation unit 113 assigns 5 points to the coordinates. By giving, the direction of the edge at the coordinates is evaluated. Further, when the deviation amount between the edge direction ⁇ and 45 ° in the coordinates is smaller than a predetermined deviation amount threshold value, the edge direction evaluation unit 113 assigns three points to the coordinates. To evaluate the edge direction at the coordinates.
- the edge direction evaluation unit 113 performs the calculation with respect to the coordinates. By not assigning points, the direction of the edge at the coordinates is evaluated. Then, the edge direction evaluation unit 113 sends data indicating the coordinates and the number of points at the coordinates to the horizontal direction distribution totaling unit 114 and the vertical direction distribution totaling unit 115.
- the deviation amount between the direction of the outline of the character and the direction of the edge that can be used for the vehicle number plate is a threshold value of a predetermined deviation amount. If it is smaller, a different score is given according to the existence rate in the direction of the contour line.
- the vertical direction of the car number plate can be more accurately determined from the image. Regions can be extracted.
- FIG. 10 shows an example of a block configuration of the image recognition apparatus 110 according to the fourth embodiment.
- the image recognition apparatus 110 according to the fourth embodiment includes an image input receiving unit 111, a gradient calculating unit 112, an edge strength determining unit 120, an edge direction evaluating unit 121, a horizontal direction distribution totaling unit 114, and a vertical direction distribution totaling unit 115.
- a vehicle number plate area extraction unit 116, a character recognition processing unit 117, and a vehicle number data transmission unit 118 In the following description, the function and operation of each component will be described in detail.
- the constituent elements of the image recognition apparatus 110 according to the fourth embodiment are the same. Functions and operations are shown. Therefore, in the following description, the detailed description is abbreviate
- the edge strength determination unit 120 determines whether the edge strength at each coordinate of the image is greater than a predetermined strength threshold value.
- the edge direction evaluation unit 121 includes a direction of an outline having a higher presence rate than a threshold of a predetermined presence rate among directions of contours of characters that can be used for a car number plate, and a direction of an edge at each coordinate of an image.
- a predetermined deviation amount threshold value points are assigned to the coordinates, and the direction of the edge at the coordinates is evaluated. For example, when the edge strength determination unit 120 determines that the edge strength is smaller than a predetermined strength threshold, the edge direction evaluation unit 121 determines the direction of the outline of the character that can be used for the vehicle number plate. Even if the amount of deviation of the coordinates from the edge direction is smaller than a predetermined deviation amount threshold value, no point is given to the coordinates.
- FIG. 11 shows an example of an operation flow of the image recognition apparatus 110 according to the fourth embodiment.
- both FIGS. 1 to 10 are referred to.
- the operation steps having the same names as the operation steps of the image recognition apparatus 110 according to the first embodiment have the same operations. Indicates. Therefore, in the following description, the detailed description is abbreviate
- the gradient calculation unit 112 of the image recognition apparatus 110 calculates a gradient value for each of the horizontal direction and the vertical direction of each coordinate, the coordinate and the horizontal gradient value f (row) of the coordinate. ) And the vertical gradient value f (col) of the coordinates are sent to the edge strength determination unit 120 and the edge direction evaluation unit 121.
- the edge strength determination unit 120 of the image recognition apparatus 110 receives the data sent from the gradient calculation unit 112, is the edge strength at the coordinates indicated by the data larger than a predetermined strength threshold value? It is determined whether or not (S111). For example, the edge strength determination unit 120 converts the horizontal gradient value f (row) of the coordinates indicated by the data received from the gradient calculation unit 112 and the vertical gradient value f (col) of the coordinates. Based on this, the edge strength at that coordinate is calculated as in equation (4) or equation (5).
- the edge strength determination unit 120 determines whether or not the calculated edge strength is greater than a predetermined strength threshold.
- the threshold value of strength is large. Set to a value.
- the strength threshold is Set to a small value.
- the edge direction evaluation unit 121 of the image recognition apparatus 110 receives the data sent from the gradient calculation unit 112 and receives the data sent from the edge direction evaluation unit 121, it is indicated by the data received from the edge direction evaluation unit 121.
- the determination result is a determination result that “the edge strength at the coordinates is smaller than a predetermined strength threshold” (S111: No)
- the direction of the outline of the character that can be used for the vehicle number plate Even if the amount of deviation of the coordinates from the edge direction is smaller than a predetermined deviation amount threshold value, points are not assigned to the coordinates (S112).
- the determination result indicated by the data received from the edge direction evaluation unit 121 is a determination result that “the edge strength at the coordinates is larger than the threshold value of the predetermined strength” (S111: Yes).
- the edge direction evaluation unit 121 evaluates the edge direction at the coordinates indicated by the data sent from the gradient calculation unit 112 (S103).
- the image recognition apparatus 110 determines whether or not the edge strength at each coordinate of the image is greater than a predetermined strength threshold value. Then, the image recognition device 110, when the edge strength determination unit determines that the edge strength is smaller than a predetermined strength threshold, the direction of the outline of the character that can be used for the vehicle number plate, Even if the amount of deviation of the coordinates from the edge direction is smaller than a predetermined deviation amount threshold value, points are not assigned to the coordinates.
- an edge at a nearby coordinate such as an uneven portion of a road surface can be ignored as noise.
- FIG. 12 shows an example of a block configuration of the image recognition apparatus 110 according to the fifth embodiment.
- the image recognition apparatus 110 according to the fifth embodiment includes an image input reception unit 111, a gradient calculation unit 112, an edge direction evaluation unit 113, a region distribution totaling unit 122, a car number plate region extraction unit 123, a character recognition processing unit 117, The vehicle number data transmission unit 118, the evaluation information storage unit 124, and the region distribution information storage unit 125 are included.
- the function and operation of each component will be described in detail.
- the components of the image recognition device 110 according to the fifth embodiment are the same. Functions and operations are shown. Therefore, in the following description, the detailed description is abbreviate
- the region distribution totaling unit 122 totalizes the distribution of the points assigned by the edge direction evaluation unit 113 for each predetermined region.
- the vehicle number plate region extraction unit 123 extracts a region of the vehicle number plate from the image based on the distribution of points given by the edge direction evaluation unit 113. For example, the vehicle number plate region extraction unit 123 extracts a region of the vehicle number plate from the image based on the distribution of points for each predetermined region that is totaled by the region distribution totalization unit 122.
- the evaluation information storage unit 124 stores information on the number of points assigned to each coordinate.
- the area distribution information storage unit 125 stores information on the distribution of points for each predetermined area.
- FIG. 13 shows an example of information stored in the evaluation information storage unit 124 in a table format.
- the evaluation information storage unit 124 stores information on coordinates and points in association with each other.
- the coordinate information is information indicating the coordinates of the target image for which the edge direction evaluation unit 113 has evaluated the direction of the edge.
- the point information is information indicating the number of points given to the coordinates indicated by the coordinate information.
- FIG. 14 shows an example of information stored in the region distribution information storage unit 125 in a table format.
- the area distribution information storage unit 125 stores information on the start point coordinates, end point coordinates, and the number of points in association with each other.
- the information of the start point coordinates indicates the upper left coordinates of the rectangular area where the area distribution totaling unit 122 totals the point distribution.
- the end point coordinate information is information indicating the lower right coordinates of the rectangular area where the area distribution totaling unit 122 has aggregated the point distribution.
- the point information is information indicating the total points of the points in the rectangular area surrounded by the coordinates indicated by the start point coordinate information and the coordinates indicated by the end point coordinate information, which are totaled by the area distribution totaling unit 122. .
- FIG. 15 shows an example of an operation flow of the image recognition apparatus 110 according to the fifth embodiment.
- both FIGS. 1 to 14 are referred to.
- the operation steps having the same names as the operation steps of the image recognition apparatus 110 according to the first embodiment have the same operations. Indicates. Therefore, in the following description, the detailed description is abbreviate
- the edge direction evaluation unit 113 of the image recognition apparatus 110 evaluates the direction of the edge at the coordinates, the information on the coordinates and the information on the number of points at the coordinates are associated with each other and stored in the evaluation information storage unit 124. In this way, information as shown in FIG. 13 is stored in the evaluation information storage unit 124.
- the edge direction evaluation unit 113 finishes evaluating the edge direction in all the coordinates of the image, the edge direction evaluation unit 113 sends data indicating that to the region distribution totaling unit 122.
- the edge direction evaluation unit 113 refers to the information stored in the evaluation information storage unit 124.
- the distribution of the score for each predetermined area is totaled (S113).
- the size of the predetermined area is a rectangular area that is the same as or smaller than the size of the car number plate on the image when the assumed car number plate is imaged.
- the area distribution totaling unit 122 adds the points in the predetermined area and totals the entire area of the image while shifting the predetermined area little by little.
- the area distribution totaling unit 122 associates the information of the start point coordinates of the predetermined area, the information of the end point coordinates of the predetermined area, and the information of the total points in the predetermined area, 125. In this way, information as shown in FIG. 14 is stored in the region distribution information storage unit 125.
- the region distribution totaling unit 122 may count the predetermined regions while shifting the pixels one pixel at a time, may count the plurality of pixels while shifting, or may total the numbers while shifting at random. Then, the area distribution totaling unit 122 sends the data indicating that to the vehicle number plate area extracting unit 123 when the distribution of the score distribution for each predetermined area is completed for all the areas of the image.
- the vehicle number plate region extraction unit 123 of the image recognition device 110 receives the data sent from the region distribution totaling unit 122, the vehicle number plate region extraction unit 123 refers to the information stored in the region distribution information storage unit 125, and The area of the number plate is extracted (S114).
- the car number plate area extraction unit 123 stores the start point stored in association with information of a score greater than a predetermined score threshold among the information stored in the region distribution information storage unit 125. All the coordinate information and end point coordinate information are read.
- the vehicle number plate region extraction unit 123 then encloses all rectangular regions surrounded by the coordinates indicated by the start point coordinate information read from the region distribution information storage unit 125 and the coordinates indicated by the end point coordinate information. Are extracted as the area of the car number plate.
- the car number plate area extracting unit 123 sends data indicating the upper left coordinates and the lower right coordinates of the extracted car number plate area to the character recognition processing unit 117.
- the image recognition apparatus 110 totals the score distribution for each predetermined area. And the image recognition apparatus 110 extracts the area
- the image recognition device 110 according to the fifth embodiment extracts the region of the vehicle number plate from the image more accurately than the image recognition device 110 according to the first embodiment. can do.
- FIG. 16 shows an example of a block configuration of the image recognition apparatus 110 according to the sixth embodiment.
- the image recognition apparatus 110 according to the sixth embodiment includes an image input receiving unit 111, a gradient calculating unit 112, an edge direction evaluating unit 113, a horizontal direction distribution totaling unit 114, a vertical direction distribution totaling unit 115, a region distribution totaling unit 122,
- the vehicle number plate region extraction unit 126, the character recognition processing unit 117, the vehicle number data transmission unit 118, the evaluation information storage unit 124, and the region distribution information storage unit 125 are included.
- the function and operation of each component will be described in detail.
- the vehicle number plate region extraction unit 126 extracts the region of the vehicle number plate from the image based on the distribution of the points given by the edge direction evaluation unit 113. For example, the vehicle number plate region extraction unit 126 extracts a horizontal region of the vehicle number plate from the image based on the distribution of the number of points in the horizontal direction counted by the horizontal direction distribution totalization unit 114. Further, for example, the vehicle number plate region extraction unit 126 extracts the vertical region of the vehicle number plate from the image based on the distribution of the vertical score calculated by the vertical direction distribution totaling unit 115. Further, for example, the vehicle number plate region extraction unit 126 extracts the region of the vehicle number plate from the image based on the distribution of points for each predetermined region that is aggregated by the region distribution aggregation unit 122.
- FIG. 17 shows an example of the operation flow of the image recognition apparatus 110 according to the sixth embodiment.
- both FIGS. 1 to 16 are referred to.
- the same reference numerals as those of the image recognition apparatus 110 according to the first embodiment or the image recognition apparatus 110 according to the fifth embodiment are used.
- the same operation steps with the same names indicate similar operations. Therefore, in the following description, the detailed description is abbreviate
- the horizontal direction distribution totaling unit 114 and the vertical direction distribution totaling unit indicate data indicating the coordinates and the number of points in the coordinates. 115 and the information is associated with each other and stored in the evaluation information storage unit 124. In this way, information as shown in FIG. 13 is stored in the evaluation information storage unit 124.
- the vehicle number plate region extraction unit 116 of the image recognition device 110 receives the data sent from the horizontal direction distribution totaling unit 114 and the vertical direction distribution totaling unit 115, respectively, based on the distribution of points indicated by these data.
- the temporary area of the car number plate is extracted from the image (S115).
- the vehicle number plate region extracting unit 116 specifies the total value Vi_max that maximizes the horizontal score distribution from the horizontal score distribution indicated by the data received from the horizontal distribution totaling unit 114. .
- the vehicle number plate region extraction unit 116 calculates a value Vi_threshold obtained by multiplying the total value Vi_max by a predetermined coefficient ⁇ (where 0 ⁇ ⁇ 1).
- the car number plate region extraction unit 116 searches the distribution of the number of points in the horizontal direction from the left end coordinates of the image to the right end direction, and as shown in FIG. 3, the aggregate value is the first value Vi_threshold of the image.
- the horizontal coordinate i_left is specified.
- the car number plate region extraction unit 116 searches the distribution of the number of points in the horizontal direction from the coordinates of the right end of the image to the left end direction, and as shown in FIG. 3, the aggregate value is the first value Vi_threshold.
- the horizontal coordinate i_right is specified.
- the vehicle number plate region extraction unit 116 specifies the total value Vj_max that maximizes the distribution of the vertical score from the distribution of the vertical score indicated by the data received from the vertical distribution calculation unit 115. To do. Then, the vehicle number plate region extraction unit 116 calculates a value Vj_threshold obtained by multiplying the total value Vj_max by a predetermined coefficient ⁇ (where 0 ⁇ ⁇ 1). Then, the car number plate area extraction unit 116 searches the distribution of the score in the vertical direction from the coordinates of the upper end of the image to the direction of the lower end, and as shown in FIG. 4, as shown in FIG. The vertical coordinate j_upper is specified.
- the car number plate region extraction unit 116 searches the distribution of the score in the vertical direction from the coordinates of the lower end of the image to the upper end direction, and as shown in FIG. The vertical coordinate j_lower is specified. Then, as shown in FIG. 5, the vehicle number plate region extraction unit 116 extracts a region surrounded by the start point coordinates (i_left, j_upper) and the end point coordinates (i_right, j_lower) as a temporary region of the vehicle number plate. Then, the vehicle number plate region extraction unit 116 sends data indicating the start point coordinates (i_left, j_upper) and end point coordinates (i_right, j_lower) of the extracted temporary region of the vehicle number plate to the region distribution totaling unit 122.
- the area distribution totaling unit 122 of the image recognition apparatus 110 receives the data sent from the vehicle number plate area extracting unit 116, the start point coordinates (i_left, j_upper) and the end point coordinates (i_right, j_lower) indicated by the data are used.
- the information stored in the evaluation information storage unit 124 is referred to, and the distribution of the points assigned by the edge direction evaluation unit 113 for each predetermined area is tabulated (S116).
- the size of the predetermined area is a rectangular area that is the same as or smaller than the size of the car number plate on the image when an assumed car number plate is imaged.
- the area distribution totaling unit 122 adds and accumulates points in the predetermined area while shifting the predetermined area little by little with respect to the temporary area of the vehicle number plate. Then, the area distribution totaling unit 122 associates the information of the start point coordinates of the predetermined area, the information of the end point coordinates of the predetermined area, and the information of the total points in the predetermined area, 125. In this way, information as shown in FIG. 14 is stored in the region distribution information storage unit 125.
- the region distribution totaling unit 122 may count the predetermined regions while shifting the pixels one pixel at a time, may count the plurality of pixels while shifting, or may total the numbers while shifting at random. And the area distribution total part 122 will send the data which show that to the vehicle number plate area extraction part 123, if the distribution of the score for every predetermined area is complete
- the vehicle number plate region extraction unit 123 of the image recognition device 110 receives the data sent from the region distribution totaling unit 122, the vehicle number plate region extraction unit 123 refers to the information stored in the region distribution information storage unit 125, and The main area of the number plate is extracted (S117).
- the car number plate area extraction unit 123 stores the start point stored in association with information of a score greater than a predetermined score threshold among the information stored in the region distribution information storage unit 125. All the coordinate information and end point coordinate information are read.
- the vehicle number plate region extraction unit 123 then encloses all rectangular regions surrounded by the coordinates indicated by the start point coordinate information read from the region distribution information storage unit 125 and the coordinates indicated by the end point coordinate information. Is extracted as the main area of the car number plate.
- the car number plate area extracting unit 123 sends data indicating the upper left coordinates and the lower right coordinates of the extracted main area of the car number plate to the character recognition processing unit 117.
- the image recognizing device 110 has the vehicle number plate out of the image based on the aggregated distribution of horizontal points and the aggregated distribution of vertical points. Extract the temporary area. Then, the image recognition device 110 extracts the main region of the vehicle number plate from the temporary region of the vehicle number plate based on the aggregated distribution for each predetermined region.
- the image recognition device 110 according to the sixth embodiment extracts the area of the vehicle number plate from the image more accurately than the image recognition device 110 according to the first embodiment.
- the area of the vehicle number plate can be extracted from the image at a higher speed.
- FIG. 18 shows an example of a hardware configuration of a computer 800 constituting the image recognition apparatus 110 according to this embodiment.
- a computer 800 includes a CPU (Central Processing Unit) 802, a RAM (Random Access Memory) 803, a graphic controller 804, and a display 805 that are mutually connected by a host controller 801, and input / output An input / output unit having a communication interface 807, a hard disk drive 808, and a CD-ROM (Compact Disk Only Memory) drive 809 connected to each other by the controller 806, and a ROM (Read Only Memory) connected to the input / output controller 806 810, legacy input / output having flexible disk drive 811 and input / output chip 812 Provided with a door.
- CPU Central Processing Unit
- RAM Random Access Memory
- FIG. 18 shows an example of a hardware configuration of a computer 800 constituting the image recognition apparatus 110 according to this embodiment.
- a computer 800 includes a CPU (Central Processing Unit) 802, a RAM (Random Access Memory) 803, a
- the host controller 801 connects the RAM 803, the CPU 802 that accesses the RAM 803 at a high transfer rate, and the graphic controller 804.
- the CPU 802 operates based on programs stored in the ROM 810 and the RAM 803 and controls each unit.
- the graphic controller 804 acquires image data generated on a frame buffer provided in the RAM 803 by the CPU 802 and the like and displays the image data on the display 805.
- the graphic controller 804 may include a frame buffer for storing image data generated by the CPU 802 or the like.
- the input / output controller 806 connects the host controller 801 to the communication interface 807, the hard disk drive 808, and the CD-ROM drive 809, which are relatively high-speed input / output devices.
- the hard disk drive 808 stores programs and data used by the CPU 802 in the computer 800.
- the CD-ROM drive 809 reads a program or data from the CD-ROM 892 and provides it to the hard disk drive 808 via the RAM 803.
- the input / output controller 806 is connected to the ROM 810, the flexible disk drive 811 and the input / output chip 812, which are relatively low-speed input / output devices.
- the ROM 810 stores a boot program that is executed when the computer 800 is started and / or a program that depends on the hardware of the computer 800.
- the flexible disk drive 811 reads a program or data from the flexible disk 893 and provides it to the hard disk drive 808 via the RAM 803.
- the input / output chip 812 connects the flexible disk drive 811 to the input / output controller 806 and connects various input / output devices to the input / output controller 806 via, for example, a parallel port, serial port, keyboard port, mouse port, and the like. To do.
- the program provided to the hard disk drive 808 via the RAM 803 is stored in a recording medium such as a flexible disk 893, a CD-ROM 892, or an IC (Integrated Circuit) card and provided by the user.
- the program is read from the recording medium, installed in the hard disk drive 808 in the computer 800 via the RAM 803, and executed by the CPU 802.
- a program that is installed in the computer 800 and causes the computer 800 to function as the image recognition device 110 causes the computer 800 to have a predetermined abundance ratio among the directions of the contour lines of characters that can be used for the vehicle number plate in steps S103, S109, and S110.
- the vehicle number plate from the image It is made to function as a vehicle number plate area extraction part which extracts the area
- the program causes the computer 800 to determine whether or not the edge strength at each coordinate of the image is greater than a predetermined strength threshold in step S111, and an edge strength determination unit. If the edge strength determination unit determines that the strength is smaller than the predetermined strength threshold, in step S112, the amount of shift between the contour direction and the edge direction is equal to the predetermined shift amount. Even if it is smaller than the threshold value, it may function as an edge direction evaluation unit that does not give points to the coordinates.
- the program causes the computer 800 to calculate the horizontal distribution of the points distributed by the edge direction evaluation unit in step S104 and the horizontal distribution of the horizontal distribution tabulated by the horizontal distribution aggregation unit. Based on the distribution, in steps S106 and S115, it may function as a vehicle number plate region extracting unit that extracts a region in the horizontal direction of the vehicle number plate from the image.
- the program causes the computer 800 to calculate the vertical direction distribution totaling unit that totals the vertical distribution of the points given by the edge direction evaluation unit in step S105 and the vertical point totals counted by the vertical direction distribution totaling unit. Based on the distribution, in steps S106 and S115, the vehicle number plate region extracting unit that extracts the region in the vertical direction of the vehicle number plate from the image may be functioned.
- the program causes the computer 800 to calculate the distribution for each predetermined area of the points given by the edge direction evaluation section in step S113 and S116, and the area distribution totaling section for each predetermined area totaled by the area distribution totaling section. Based on the distribution of the points, in steps S114 and S117, a vehicle number plate region extracting unit that extracts a region of the vehicle number plate from the image may be functioned.
- the program You may make it function as an edge direction evaluation part which provides a different score according to the presence rate of the direction of an outline.
- the information processing described in these programs is read into the computer 800, whereby the image input receiving unit, the gradient calculating unit, the edge direction evaluation, which are specific means in which the software and the various hardware resources described above cooperate.
- the specific image recognition apparatus 110 according to the use purpose is constructed
- the CPU 802 executes a communication program loaded on the RAM 803 and executes a communication interface based on the processing content described in the communication program.
- a communication process is instructed to 807.
- the communication interface 807 reads transmission data stored in a transmission buffer area or the like provided on a storage device such as the RAM 803, the hard disk drive 808, the flexible disk 893, or the CD-ROM 892, and sends it to the network.
- the reception data transmitted or received from the network is written into a reception buffer area or the like provided on the storage device.
- the communication interface 807 may transfer transmission / reception data to / from the storage device by the direct memory access method. Instead, the CPU 802 reads data from the transfer source storage device or the communication interface 807.
- the transmission / reception data may be transferred by writing the data to the transfer destination communication interface 807 or the storage device.
- the CPU 802 transfers all or a necessary portion of the files or database stored in the external storage device such as the hard disk drive 808, CD-ROM 892, and flexible disk 893 to the RAM 803 by direct memory access transfer or the like.
- the data is read and various processes are performed on the data on the RAM 803. Then, the CPU 802 writes the processed data back to the external storage device by direct memory access transfer or the like.
- the RAM 803 can be regarded as temporarily holding the contents of the external storage device, in the present embodiment, the RAM 803 and the external storage device are collectively referred to as a memory, a storage unit, or a storage device. .
- Various types of information such as various programs, data, tables, and databases in the present embodiment are stored on such a storage device and are subjected to information processing.
- the CPU 802 can hold a part of the RAM 803 in the cache memory and perform reading and writing on the cache memory. Even in such a form, the cache memory bears a part of the function of the RAM 803. Therefore, in the present embodiment, the cache memory is also included in the RAM 803, the memory, and / or the storage device unless otherwise indicated. To do.
- the CPU 802 performs various operations, such as various operations, information processing, condition determination, information retrieval, replacement, and the like described in the present embodiment for data read from the RAM 803 and specified by a command sequence of the program. Is written back to the RAM 803. For example, when performing the condition determination, the CPU 802 satisfies the conditions such that the various variables shown in the present embodiment are larger, smaller, above, below, or equal to other variables or constants. If the condition is satisfied or not satisfied, the program branches to a different instruction sequence or calls a subroutine.
- the CPU 802 can search for information stored in a file in a storage device or a database. For example, when a plurality of entries in which the attribute value of the second attribute is associated with the attribute value of the first attribute are stored in the storage device, the CPU 802 stores the plurality of entries stored in the storage device. The entry that matches the condition in which the attribute value of the first attribute is specified is retrieved, and the attribute value of the second attribute that is stored in the entry is read, thereby associating with the first attribute that satisfies the predetermined condition The attribute value of the specified second attribute can be obtained.
- the programs or modules shown above may be stored in an external storage medium.
- a storage medium in addition to a flexible disk 893 and a CD-ROM 892, an optical recording medium such as a DVD (Digital Versatile Disk) or a CD (Compact Disk), a magneto-optical recording medium such as an MO (Magneto-Optical disk), or a tape
- a storage medium such as a hard disk or RAM provided in a server system connected to a dedicated communication network or the Internet may be used as a recording medium, and the program may be provided to the computer 800 via the network.
- the region of the car number plate can be extracted from the image.
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
- Traffic Control Systems (AREA)
- Character Input (AREA)
Abstract
Description
本願は、2012年11月16日に、日本に出願された特願2012-251806号に基づき優先権を主張し、その内容をここに援用する。
また、これらの特徴群のサブコンビネーションもまた、発明となり得る。
車番プレート領域抽出部116が抽出した車番プレートの領域は、始点座標(i_left,j_upper)、及び終点座標(i_right,j_lower)で囲まれる領域である。
一方、座標におけるエッジの方向θと、0°、45°、又は90°とのずれ量が、所定のずれ量のしきい値よりも大きい場合、エッジ方向評価部113は、その座標に対して点数を付与しないことによって、その座標におけるエッジの方向を評価する。そして、エッジ方向評価部113は、その座標と、その座標における点数とを示すデータを、水平方向分布集計部114、及び鉛直方向分布集計部115へ送る。
そのため、路面の凹凸部分等の近傍の座標におけるエッジは、ノイズとして無視する必要がある。そこで、エッジ強さ判定部120は、算出したエッジの強さが所定の強さのしきい値よりも大きいか否かを判定する。例えば、車番プレートの文字の輪郭線部分の画素値が車番プレート本体の画素値と比較して、大きく変化しているようなものを対象とする場合、強さのしきい値は、大きい値に設定される。一方、車番プレートの文字の輪郭線部分の画素値が車番プレート本体の画素値と比較して、あまり大きく変化していないようなものを対象とする場合、強さのしきい値は、小さい値に設定される。そして、エッジ強さ判定部120は、座標におけるエッジの強さが所定の強さのしきい値よりも大きいか否かを判定すると、その座標と、判定結果とを示すデータを、エッジ方向評価部121へ送る。
領域分布情報格納部125には、始点座標、終点座標、及び点数の各情報が対応付けられて格納される。
そして、車番プレート領域抽出部116は、抽出した車番プレートの仮領域の始点座標(i_left,j_upper)、及び終点座標(i_right,j_lower)を示すデータを、領域分布集計部122へ送る。
110 画像認識装置
111 画像入力受付部
112 勾配算出部
113 エッジ方向評価部
114 水平方向分布集計部
115 鉛直方向分布集計部
116 車番プレート領域抽出部
117 文字認識処理部
118 車番データ送信部
119 ルックアップテーブル
120 エッジ強さ判定部
121 エッジ方向評価部
122 領域分布集計部
123 車番プレート領域抽出部
124 評価情報格納部
125 領域分布情報格納部
126 車番プレート領域抽出部
130 CCTVカメラ
150 補助光源
170 ホストコンピュータ
800 コンピュータ
801 ホストコントローラ
802 CPU
803 RAM
804 グラフィックコントローラ
805 ディスプレイ
806 入出力コントローラ
807 通信インターフェース
808 ハードディスクドライブ
809 CD-ROMドライブ
810 ROM
811 フレキシブルディスクドライブ
812 入出力チップ
891 ネットワーク通信装置
892 CD-ROM
893 フレキシブルディスク
N 通信回線
Claims (9)
- 画像の中から車番プレートの領域を抽出する画像認識装置であって、
前記車番プレートに使用され得る文字の輪郭線の方向のうち所定の存在率のしきい値よりも高い存在率の輪郭線の方向と、前記画像の各座標におけるエッジの方向とのずれ量が、所定のずれ量のしきい値よりも小さい場合に、前記座標に対して点数を付与して前記座標におけるエッジの方向を評価するエッジ方向評価部と、
前記エッジ方向評価部が付与した点数の分布に基づいて、前記画像の中から前記車番プレートの領域を抽出する車番プレート領域抽出部と
を備える画像認識装置。 - 前記画像の各座標におけるエッジの強さが所定の強さのしきい値よりも大きいか否かを判定するエッジ強さ判定部
を更に備え、
前記エッジ方向評価部は、前記エッジの強さが所定の強さのしきい値よりも小さいと前記エッジ強さ判定部が判定した場合、前記輪郭線の方向と、前記エッジの方向とのずれ量が、所定のずれ量のしきい値よりも小さくても、前記座標に対して点数を付与しない
請求項1に記載の画像認識装置。 - 前記エッジ方向評価部が付与した点数の水平方向の分布を集計する水平方向分布集計部
を更に備え、
前記車番プレート領域抽出部は、前記水平方向分布集計部が集計した水平方向の点数の分布に基づいて、前記画像の中から前記車番プレートの水平方向の領域を抽出する
請求項1又は2に記載の画像認識装置。 - 前記エッジ方向評価部が付与した点数の鉛直方向の分布を集計する鉛直方向分布集計部
を更に備え、
前記車番プレート領域抽出部は、前記鉛直方向分布集計部が集計した鉛直方向の点数の分布に基づいて、前記画像の中から前記車番プレートの鉛直方向の領域を抽出する
請求項1から3のいずれか一項に記載の画像認識装置。 - 前記エッジ方向評価部が付与した点数の所定領域毎の分布を集計する領域分布集計部
を更に備え、
前記車番プレート領域抽出部は、前記領域分布集計部が集計した所定領域毎の点数の分布に基づいて、前記画像の中から前記車番プレートの領域を抽出する
請求項1から4のいずれか一項に記載の画像認識装置。 - 前記エッジ方向評価部は、前記輪郭線の方向と前記エッジの方向とのずれ量が所定のずれ量のしきい値よりも小さい場合、前記輪郭線の方向の存在率に応じて異なる点数を付与する
請求項1から5のいずれか一項に記載の画像認識装置。 - 画像の中から車番プレートの領域を抽出する車番プレート領域抽出方法であって、
前記車番プレートに使用され得る文字の輪郭線の方向のうち所定の存在率のしきい値よりも高い存在率の輪郭線の方向と、前記画像の各座標におけるエッジの方向とのずれ量が、所定のずれ量のしきい値よりも小さい場合に、前記座標に対して点数を付与して前記座標におけるエッジの方向を評価するエッジ方向評価段階と、
前記エッジ方向評価段階において付与された点数の分布に基づいて、前記画像の中から前記車番プレートの領域を抽出する車番プレート領域抽出段階と
を備える画像認識方法。 - 画像の中から車番プレートの領域を抽出する画像認識装置として、コンピュータを機能させるプログラムであって、
前記コンピュータを、
前記車番プレートに使用され得る文字の輪郭線の方向のうち所定の存在率のしきい値よりも高い存在率の輪郭線の方向と、前記画像の各座標におけるエッジの方向とのずれ量が、所定のずれ量のしきい値よりも小さい場合に、前記座標に対して点数を付与して前記座標におけるエッジの方向を評価するエッジ方向評価部、
前記エッジ方向評価部が付与した点数の分布に基づいて、前記画像の中から前記車番プレートの領域を抽出する車番プレート領域抽出部
として機能させるプログラム。 - 画像の中から車番プレートの領域を抽出する画像認識装置として、コンピュータを機能させるプログラムを記録した記録媒体であって、
前記コンピュータを、
前記車番プレートに使用され得る文字の輪郭線の方向のうち所定の存在率のしきい値よりも高い存在率の輪郭線の方向と、前記画像の各座標におけるエッジの方向とのずれ量が、所定のずれ量のしきい値よりも小さい場合に、前記座標に対して点数を付与して前記座標におけるエッジの方向を評価するエッジ方向評価部、
前記エッジ方向評価部が付与した点数の分布に基づいて、前記画像の中から前記車番プレートの領域を抽出する車番プレート領域抽出部
として機能させるプログラムを記録した記録媒体。
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| SG11201503823UA SG11201503823UA (en) | 2012-11-16 | 2013-03-27 | Image recognition device, image recognition method, program, and recording medium |
| US14/442,715 US9747511B2 (en) | 2012-11-16 | 2013-03-27 | Image recognition device, image recognition method, program, and recording medium |
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP2012251806A JP6037791B2 (ja) | 2012-11-16 | 2012-11-16 | 画像認識装置、画像認識方法、プログラム、及び記録媒体 |
| JP2012-251806 | 2012-11-16 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2014076979A1 true WO2014076979A1 (ja) | 2014-05-22 |
Family
ID=50730908
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/JP2013/059061 Ceased WO2014076979A1 (ja) | 2012-11-16 | 2013-03-27 | 画像認識装置、画像認識方法、プログラム、及び記録媒体 |
Country Status (6)
| Country | Link |
|---|---|
| US (1) | US9747511B2 (ja) |
| JP (1) | JP6037791B2 (ja) |
| MY (1) | MY169870A (ja) |
| SG (1) | SG11201503823UA (ja) |
| TW (1) | TWI482103B (ja) |
| WO (1) | WO2014076979A1 (ja) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110348440A (zh) * | 2019-07-09 | 2019-10-18 | 北京字节跳动网络技术有限公司 | 牌照检测方法、装置、电子设备及存储介质 |
| CN112885134A (zh) * | 2021-01-24 | 2021-06-01 | 成都智慧赋能科技有限公司 | 一种基于大数据的智慧城市交通管理方法 |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| IL238473A0 (en) * | 2015-04-26 | 2015-11-30 | Parkam Israel Ltd | A method and system for discovering and mapping parking areas |
| CN108320582B (zh) * | 2018-03-30 | 2020-01-24 | 合肥城市泊车投资管理有限公司 | 一种具备剩余车位统计功能的停车管理系统 |
| JP7467107B2 (ja) * | 2019-12-25 | 2024-04-15 | キヤノン株式会社 | 画像処理装置、画像処理方法、およびプログラム |
| CN114863128A (zh) * | 2022-03-23 | 2022-08-05 | 佛山科学技术学院 | 一种涂胶单板的轮廓提取和纠偏系统及方法 |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPS6488210A (en) * | 1987-09-30 | 1989-04-03 | Suzuki Motor Co | Method for detecting distance between cars |
| JPH04169987A (ja) * | 1990-11-01 | 1992-06-17 | Omron Corp | 文字列検出方法、文字列検出装置、文字列読取方法及び文字列読取装置 |
| JPH08329252A (ja) * | 1995-05-31 | 1996-12-13 | Sony Corp | エッジ検出方法及びエッジ検出装置 |
| JPH11339017A (ja) * | 1998-05-29 | 1999-12-10 | Mitsubishi Heavy Ind Ltd | ナンバープレート認識装置 |
Family Cites Families (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US647351A (en) * | 1899-08-24 | 1900-04-10 | John Chandler Wood | Valve. |
| JP2893814B2 (ja) | 1990-03-15 | 1999-05-24 | 住友電気工業株式会社 | 車番自動読取装置におけるプレート切出し装置 |
| US6281928B1 (en) * | 1998-05-13 | 2001-08-28 | Chuo Hatsujo Kabushiki Kaisha | Positional detector device for a vehicular license plate |
| JP3679958B2 (ja) | 1999-10-28 | 2005-08-03 | 三菱重工業株式会社 | ナンバープレート認識装置 |
| TWI262454B (en) * | 2002-04-22 | 2006-09-21 | Chunghwa Telecom Co Ltd | Image-type plate recognizer |
| TWI236638B (en) * | 2004-03-24 | 2005-07-21 | Univ Chung Yuan Christian | Multiple recognition system and method for license plates |
| IL162878A0 (en) * | 2004-07-06 | 2005-11-20 | Hi Tech Solutions Ltd | Multi-level neural network based characters identification method and system |
| JP4713107B2 (ja) * | 2004-08-20 | 2011-06-29 | 日立オムロンターミナルソリューションズ株式会社 | 景観中文字列認識方式および装置 |
| TW200802137A (en) * | 2006-06-16 | 2008-01-01 | Univ Nat Chiao Tung | Serial-type license plate recognition system |
| WO2008079902A1 (en) * | 2006-12-21 | 2008-07-03 | Neology, Inc. | Systems and methods for a rfid enabled metal license plate |
| CN101458814B (zh) * | 2007-12-13 | 2012-02-01 | 东软集团股份有限公司 | 一种用于从图像中分割对象感兴趣区域的方法和装置 |
| TWI430187B (zh) * | 2011-02-14 | 2014-03-11 | Chunghwa Telecom Co Ltd | License plate number identification method |
| US20140254878A1 (en) * | 2013-03-08 | 2014-09-11 | Next Level Security Systems, Inc. | System and method for scanning vehicle license plates |
| US20140254866A1 (en) * | 2013-03-08 | 2014-09-11 | Next Level Security Systems, Inc. | Predictive analysis using vehicle license plate recognition |
| US9460367B2 (en) * | 2014-11-26 | 2016-10-04 | Xerox Corporation | Method and system for automating an image rejection process |
-
2012
- 2012-11-16 JP JP2012251806A patent/JP6037791B2/ja active Active
-
2013
- 2013-03-27 US US14/442,715 patent/US9747511B2/en active Active
- 2013-03-27 WO PCT/JP2013/059061 patent/WO2014076979A1/ja not_active Ceased
- 2013-03-27 MY MYPI2015701493A patent/MY169870A/en unknown
- 2013-03-27 SG SG11201503823UA patent/SG11201503823UA/en unknown
- 2013-03-27 TW TW102111009A patent/TWI482103B/zh not_active IP Right Cessation
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPS6488210A (en) * | 1987-09-30 | 1989-04-03 | Suzuki Motor Co | Method for detecting distance between cars |
| JPH04169987A (ja) * | 1990-11-01 | 1992-06-17 | Omron Corp | 文字列検出方法、文字列検出装置、文字列読取方法及び文字列読取装置 |
| JPH08329252A (ja) * | 1995-05-31 | 1996-12-13 | Sony Corp | エッジ検出方法及びエッジ検出装置 |
| JPH11339017A (ja) * | 1998-05-29 | 1999-12-10 | Mitsubishi Heavy Ind Ltd | ナンバープレート認識装置 |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110348440A (zh) * | 2019-07-09 | 2019-10-18 | 北京字节跳动网络技术有限公司 | 牌照检测方法、装置、电子设备及存储介质 |
| CN112885134A (zh) * | 2021-01-24 | 2021-06-01 | 成都智慧赋能科技有限公司 | 一种基于大数据的智慧城市交通管理方法 |
Also Published As
| Publication number | Publication date |
|---|---|
| SG11201503823UA (en) | 2015-06-29 |
| TW201421379A (zh) | 2014-06-01 |
| US9747511B2 (en) | 2017-08-29 |
| MY169870A (en) | 2019-05-27 |
| JP2014099128A (ja) | 2014-05-29 |
| JP6037791B2 (ja) | 2016-12-07 |
| US20160132742A1 (en) | 2016-05-12 |
| TWI482103B (zh) | 2015-04-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP6037791B2 (ja) | 画像認識装置、画像認識方法、プログラム、及び記録媒体 | |
| CN111742344B (zh) | 图像语义分割方法、可移动平台及存储介质 | |
| US9002066B2 (en) | Methods, systems and processor-readable media for designing a license plate overlay decal having infrared annotation marks | |
| CN108364010A (zh) | 一种车牌识别方法、装置、设备及计算机可读存储介质 | |
| CN114119991B (zh) | 一种目标检测的方法、装置、电子设备及存储介质 | |
| JP4517003B2 (ja) | 道路標識自動認識システム | |
| CN111783573B (zh) | 一种远光灯检测方法、装置及设备 | |
| CN112330646A (zh) | 一种基于二维图像的动车车底异常检测方法 | |
| CN112580734A (zh) | 目标检测模型训练方法、系统、终端设备及存储介质 | |
| CN115661522B (zh) | 一种基于视觉语义矢量的车辆导引方法、系统、设备和介质 | |
| CN112926461B (zh) | 神经网络训练、行驶控制方法及装置 | |
| CN113490948A (zh) | 轮胎侧壁成像方法 | |
| KR102403169B1 (ko) | 이미지 분석을 통한 가이드 제공 방법 및 이를 실행시키기 위하여 기록매체에 기록된 컴퓨터 프로그램 | |
| CN114511832B (zh) | 车道线分析方法、装置、电子设备及存储介质 | |
| KR102389998B1 (ko) | 비식별 처리 방법 및 이를 실행시키기 위하여 기록매체에 기록된 컴퓨터 프로그램 | |
| CN113449629B (zh) | 基于行车视频的车道线虚实识别装置、方法、设备及介质 | |
| WO2013141034A1 (ja) | 車両データ処理システム、車両データ処理方法、車両データ処理装置、プログラム、及び記録媒体 | |
| CN106980859A (zh) | 一种用于识别铝车轮背腔字符的方法 | |
| CN108537223A (zh) | 一种车牌检测方法、系统及设备和存储介质 | |
| CN111161542B (zh) | 车辆识别方法及装置 | |
| CN112633183B (zh) | 影像遮挡区域自动检测方法、装置及存储介质 | |
| CN110135288A (zh) | 一种电子证照的快速核对方法及装置 | |
| KR102546193B1 (ko) | 컬러 정보를 이용한 학습 데이터 분류 방법 및 이를 실행하기 위하여 기록매체에 기록된 컴퓨터 프로그램 | |
| KR102546195B1 (ko) | 수집 환경 기반 학습 데이터 분류 방법 및 이를 실행하기 위하여 기록매체에 기록된 컴퓨터 프로그램 | |
| KR102599196B1 (ko) | 학습 데이터 생성에 관한 작업 비용 예측 방법 및 이를 실행하기 위하여 기록매체에 기록된 컴퓨터 프로그램 |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 13855334 Country of ref document: EP Kind code of ref document: A1 |
|
| WWE | Wipo information: entry into national phase |
Ref document number: IDP00201502890 Country of ref document: ID Ref document number: 14442715 Country of ref document: US |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |
|
| 122 | Ep: pct application non-entry in european phase |
Ref document number: 13855334 Country of ref document: EP Kind code of ref document: A1 |




