CN108776792B - Multi-scale positioning fusion method and device for license plate - Google Patents
Multi-scale positioning fusion method and device for license plate Download PDFInfo
- Publication number
- CN108776792B CN108776792B CN201810582289.0A CN201810582289A CN108776792B CN 108776792 B CN108776792 B CN 108776792B CN 201810582289 A CN201810582289 A CN 201810582289A CN 108776792 B CN108776792 B CN 108776792B
- Authority
- CN
- China
- Prior art keywords
- license plate
- area
- queue
- sor
- plate area
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/22—Image preprocessing by selection of a specific region containing or referencing a pattern; Locating or processing of specific regions to guide the detection or recognition
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention provides a multiscale positioning fusion method of a license plate, which comprises the following steps: inputting or collecting a gray level image; carrying out multi-scale processing on the gray level image to obtain a multi-scale image; carrying out coarse positioning on the license plate area of the multi-scale image to obtain a first license plate area sequence; filtering the first license plate region sequence according to a false license plate filtering method; accurately positioning the first license plate region sequence, restoring the coordinates to the original resolution, acquiring a second license plate region queue, searching an overlapping region of the second license plate region sequence to acquire a third license plate region queue, and finely positioning and filtering the third license plate region queue in groups to acquire a fourth license plate region queue; and (4) filtering the contact ratio of the fourth license plate queue to obtain a filtered license plate area and outputting the filtered license plate area. Compared with the existing license plate recognition technology, the width recognition range of the license plate can be effectively expanded.
Description
Technical Field
The invention relates to image processing, video monitoring and intelligent traffic, in particular to a license plate positioning method and device.
Background
With the increasingly modern city, the holding capacity of motor vehicles is continuously increased, and the intelligent traffic system plays an increasingly important role. The license plate recognition technology is an important component of a modern intelligent traffic system and has very wide application.
The license plate recognition technology mainly comprises the following steps: the method comprises three parts of license plate positioning, license plate segmentation and license plate character recognition, wherein accurate license plate positioning is a premise and a basis for correct license plate recognition. Due to the complex application environment of the license plate recognition technology, such as complex road environment, different illumination conditions, the pollution of the license plate, numerous license plate types and the like, in order to realize the low missing report rate and false report rate of the license plate positioning and simultaneously require as little time as possible, the current common license plate recognition system only supports the license plate recognition with the width of 60-200 pixels.
However, with the increasing resolution of image acquisition systems and the increasing expansion of focal length, it is becoming more and more urgent to increase the license plate width recognition range.
In summary, there is an urgent need to provide a license plate location method for improving the license plate identification width range.
Disclosure of Invention
In view of this, the main objective of the present invention is to achieve license plate location and improve the width range of license plate recognition.
To achieve the above object, according to a first aspect of the present invention, there is provided a multi-scale localization fusion method for a license plate, the method comprising:
the method comprises the steps of firstly, inputting or collecting a gray image;
the second step, carry on the multi-scale processing to the gray level picture, obtain the multi-scale picture;
thirdly, carrying out coarse positioning on the license plate area of the multi-scale image to obtain a first license plate area queue;
fourthly, filtering the first license plate area queue according to a false license plate filtering method;
a fifth step of accurately positioning the first license plate area queue and reducing the coordinates to the original resolution to obtain a second license plate area queue, searching the overlapping area of the second license plate area queue to obtain a third license plate area queue, and finely positioning and filtering the third license plate area queue in groups to obtain a fourth license plate area queue;
and sixthly, filtering the contact ratio of the fourth license plate area queue to obtain a filtered license plate area and outputting the filtered license plate area.
Further, the second step includes:
a scaling frequency calculation step, namely taking the gray level image as an original resolution image, and respectively calculating the compression frequency and the amplification frequency according to the license plate width recognition range of the original resolution image and the expanded license plate width recognition range requirement;
a low resolution image acquisition step of extracting or compressing the original resolution image according to the compression times to acquire a low resolution image;
a high-resolution image acquisition step of performing interpolation or amplification processing on the original high-resolution image according to the amplification times to acquire a high-resolution image;
and a multi-scale image output step of outputting the original resolution image, the low resolution image and the high resolution image as multi-scale images.
Further, the scaling number calculation step includes: the license plate width recognition range for the original resolution image is [ WR1，WR2]The requirement of the expanded license plate width recognition range is [ EWR1，EWR2]Calculating
Taking the integer as the amplification number, and calculating
And taking an integer as the compression time, and n is a scaling exponent.
Further, the fifth step includes:
a license plate fine positioning step, wherein the license plates of all license plate areas in the first license plate area queue are finely positioned on the corresponding scale image, the average heights of a license plate fine positioning frame and license plate character blocks are obtained, the coordinates of the license plate fine positioning frame and the average heights of the license plate character blocks are restored to the original resolution, and then the license plate fine positioning frame and the license plate character blocks are placed into a second license plate area queue rectB;
an overlapping area searching step, namely searching the overlapping area of the license plate area in the second license plate area queue rectB, and putting the coordinates of the overlapping area and the average height of the character block into a third license plate area queue rectC;
grouping calculation, namely calculating the circumscribed rectangular frames RectT of all license plate areas in the third license plate area queue rectC; counting the upper limit UL of the license plate region identification range of the ith scale imageiAnd lower limit DLiIf the average height of the character block is located at λ × ULiAnd λ × DLiClassifying the license plate region into the ith scale image; counting the number of the classified license plate regions in each scale image;
and (3) a license plate fine positioning filtering step, wherein if the number of the classified license plate regions in the ith scale image is greater than 0, the license plate fine positioning is carried out on the circumscribed rectangular frame RectT of the scale image, the average height of the character blocks of the finely positioned license plate region in the scale image is counted, and if the average height is located at lambda multiplied by ULiAnd λ × DLiOtherwise, filtering out the corresponding fine positioning license plate area;
and a fourth license plate area queue obtaining step, namely restoring the coordinates of the fine positioning license plate area of the scale image to the original resolution ratio, and then placing the coordinates into a fourth license plate area queue rectD.
Further, the sixth step includes:
a coincidence proportion calculation step, namely respectively counting two random license plate regions PRect in the fourth license plate region queuepAnd selectqOverlap region select ofpqCalculating the coincidence region PRectpqRespectively in the license plate regionpAnd selectqOccupied area ratio SORpq/SRp、SORpq/SRqWherein SRp、SRq、SORpqRespectively a license plate region PRectpAnd selectqAnd a superposed region selectpqThe area of (d);
coincidence of filtering step if SORpq/SRpAnd SORpq/SRqIf the number of the license plate is not less than the fourth threshold value, the license plate area PRect is reservedpAnd selectqFiltering out the license plate region with larger area from the license plate region with smaller area; if SORpq/SRpAnd SORpq/SRqIs not less than the fourth threshold value, the SOR is retainedpq/SRpAnd SORpq/SRqThe number plate area corresponding to the fourth threshold value is not less than the middle threshold value, and the SOR is filteredpq/SRpAnd SORpq/SRqThe number plate area is smaller than the number plate area corresponding to the fourth threshold value; if SORpq/SRpAnd SORpq/SRqAre all less than the fourth threshold value, the license plate area PRect is reserved at the same timepAnd selectq；
And a result output step, namely taking the license plate area reserved in the fourth license plate area queue as a result and outputting the result.
According to another aspect of the present invention, there is provided a multi-scale localization fusion apparatus for a license plate, the apparatus comprising:
the gray level image input or acquisition module is used for inputting or acquiring a gray level image;
the multi-scale image acquisition module is used for carrying out multi-scale processing on the gray level image to acquire a multi-scale image;
the license plate coarse positioning module is used for performing license plate area coarse positioning on the multi-scale image to obtain a first license plate area queue;
the false license plate filtering module is used for filtering the first license plate area queue according to a false license plate filtering method;
the license plate fine positioning and filtering module is used for accurately positioning the first license plate area queue, restoring the coordinates to the original resolution, acquiring a second license plate area queue, searching an overlapping area of the second license plate area queue to acquire a third license plate area queue, and performing grouping fine positioning and filtering on the third license plate area queue to acquire a fourth license plate area queue;
and the license plate contact ratio filtering module is used for filtering the contact ratio of the fourth license plate region queue to obtain a filtered license plate region and outputting the filtered license plate region.
Further, the multi-scale image acquisition module comprises:
the scaling frequency calculation module is used for taking the gray level image as an original resolution image, and respectively calculating the compression frequency and the amplification frequency according to the license plate width recognition range of the original resolution image and the expanded license plate width recognition range requirement;
the low-resolution image acquisition module is used for extracting or compressing the original resolution image according to the compression times to acquire a low-resolution image;
the high-resolution image acquisition module is used for carrying out interpolation or amplification processing on the original high-resolution image according to the amplification times to acquire a high-resolution image;
and the multi-scale image output module is used for outputting the original resolution image, the low resolution image and the high resolution image as multi-scale images.
Further, the scaling number calculation module comprises: the license plate width recognition range for the original resolution image is [ WR1,WR2]The requirement of the expanded license plate width recognition range is [ EWR1,EWR2]Calculating
Taking the integer as the amplification number, and calculating
And taking an integer as the compression time, and n is a scaling exponent.
Further, the license plate fine positioning and filtering module comprises:
the license plate fine positioning module is used for carrying out fine positioning on license plates on the images with the corresponding scales on all license plate areas in the first license plate area queue to obtain the average heights of a license plate fine positioning frame and license plate character blocks, reducing the coordinates of the license plate fine positioning frame and the average heights of the license plate character blocks to the original resolution, and then placing the license plate fine positioning frame and the license plate character blocks into a second license plate area queue rectB;
the overlapping area searching module is used for searching the overlapping area of the license plate area in the second license plate area queue rectB and putting the coordinates of the overlapping area and the average height of the character block into a third license plate area queue rectC;
the grouping calculation module is used for calculating the circumscribed rectangular frames RectT of all license plate areas in the third license plate area queue rectC; counting the upper limit UL of the license plate region identification range of the ith scale imageiAnd lower limit DLiIf the average height of the character block is located at λ × ULiAnd λ × DLiClassifying the license plate region into the ith scale image; counting the number of the classified license plate regions in each scale image;
a license plate fine positioning filtering module, configured to perform license plate fine positioning on an external rectangular frame RectT of the scale image if the number of classified license plate regions in the ith scale image is greater than 0, count an average height of character blocks in the fine positioning license plate region in the scale image, and if the average height is located at λ × ULiAnd λ × DLiOtherwise, filtering out the corresponding fine positioning license plate area;
and the fourth license plate area queue acquisition module is used for restoring the coordinates of the fine positioning license plate area of the scale image to the original resolution ratio and then placing the coordinates into a fourth license plate area queue rectD.
Further, the license plate contact ratio filtering module comprises:
a coincidence proportion calculation module for respectively counting two arbitrary license plate regions PRect in the fourth license plate region queuepAnd selectqOverlap region select ofpqCalculating the coincidence region PRectpqRespectively in the license plate regionpAnd selectqOccupied area ratio SORpq/SRp、SORpq/SRqWherein SRp、SRq、SORpqRespectively a license plate region PRectpAnd selectqAnd a superposed region selectpqThe area of (d);
coincidence filter module for if SORpq/SRpAnd SORpq/SRqIf the number of the license plate is not less than the fourth threshold value, the license plate area PRect is reservedpAnd selectqFiltering out the license plate region with larger area from the license plate region with smaller area; if SORpq/SRpAnd SORpq/SRqIs not less than the fourth threshold value, the SOR is retainedpq/SRpAnd SORpq/SRqThe number plate area corresponding to the fourth threshold value is not less than the middle threshold value, and the SOR is filteredpq/SRpAnd SORpq/SRqThe number plate area is smaller than the number plate area corresponding to the fourth threshold value; if SORpq/SRpAnd SORpq/SRqAre all less than the fourth threshold value, the license plate area PRect is reserved at the same timepAnd selectq；
And the result output module is used for outputting the license plate area reserved in the fourth license plate area queue as a result.
Compared with the existing license plate positioning technology, the multi-scale positioning fusion method and device for the license plate can quickly position the license plate and effectively improve the width recognition range of the license plate.
Drawings
Fig. 1 shows a flow chart of a multi-scale localization fusion method of a license plate according to the present invention.
Fig. 2 shows a frame diagram of a multi-scale localization fusion apparatus for license plates according to the present invention.
Detailed Description
To further clarify the structure, characteristics and other objects of the present invention, those skilled in the art will now describe in detail the preferred embodiments of the present invention with reference to the attached drawings, which are provided for the purpose of describing the technical solutions of the present invention only and are not intended to limit the present invention.
FIG. 1 is a flow chart of a multi-scale positioning fusion method for a license plate according to the present invention. As shown in fig. 1, the multi-scale positioning fusion method for license plates according to the present invention includes:
a first step S1 of inputting or acquiring a grayscale image;
a second step S2, performing multi-scale processing on the gray level image to obtain a multi-scale image;
step S3, coarse positioning of license plate areas is carried out on the multi-scale images, and a first license plate area queue is obtained;
a fourth step S4, filtering the first license plate area queue according to a false license plate filtering method;
a fifth step S5, accurately positioning the first license plate area queue, restoring the coordinates to the original resolution, acquiring a second license plate area queue, searching the overlapping area of the second license plate area queue to acquire a third license plate area queue, and finely positioning and filtering the third license plate area queue in groups to acquire a fourth license plate area queue;
and a sixth step S6, filtering the coincidence degree of the fourth license plate area queue to obtain a filtered license plate area and outputting the filtered license plate area.
The first step S1 may be to acquire a grayscale image through an image collector, or to directly input the grayscale image.
Further, the first step S1 may also be implemented by inputting or acquiring a color image, and performing a graying process on the color image to obtain a grayscale image.
Further, the second step S2 includes:
a scaling frequency calculation step S21, wherein the gray level image is used as an original resolution image, and the compression frequency and the scaling frequency are respectively calculated according to the license plate width recognition range of the original resolution image and the requirement of the expanded license plate width recognition range;
a low-resolution image acquisition step S22 of extracting or compressing the original resolution image according to the number of compression times to acquire a low-resolution image;
a high-resolution image acquisition step S23 of interpolating or enlarging the original high-resolution image according to the number of times of enlargement to acquire a high-resolution image;
the multi-scale image output step S24 outputs the original resolution image, the low resolution image, and the high resolution image as multi-scale images.
Further, the scaling number calculation step S21 includes: the license plate width recognition range for the original resolution image is [ WR1,WR2]The requirement of the expanded license plate width recognition range is [ EWR1,EWR2]Calculating
Taking the integer as the amplification number, and calculating
And taking an integer as the compression time, and n is a scaling exponent.
Further, the value range of n is 2-4.
In the first embodiment, if the license plate width recognition range of the original resolution image is [60,200], the requirement of the expanded license plate width recognition range is [60,400], and n is 2, the number of compression is 1, and the number of amplification is 0; performing compression processing on the original resolution image for 1 time to obtain a low-resolution image with a license plate width identification range of [120,400 ]; the original image with high resolution is not amplified, and no image with high resolution exists; the output multi-scale image is: original resolution image, low resolution image. In the second embodiment, if the license plate width recognition range of the original resolution image is [60,200], the requirement of the expanded license plate width recognition range is [30,600], and n is 2, the number of compression is 2, and the number of amplification is 1; performing compression processing on the original resolution image for 2 times to obtain a first low-resolution image with a license plate width identification range of [120,400] and a second low-resolution image with a license plate width identification range of [240,800 ]; carrying out 1-time amplification processing on the original resolution image to obtain a high-resolution image with a license plate width identification range of [30,100 ]; the scale image output is: the image processing device comprises an original resolution image, a first low resolution image, a second low resolution image and a high resolution image.
Further, the third step S3 includes: and carrying out coarse positioning on the license plate area of the ith scale image to obtain the coordinates of the license plate area, and putting the coordinates into a first license plate area queue rectA [ i ].
The coarse positioning in the third step S3 can be implemented by using an existing license plate coarse positioning algorithm.
In the embodiment, for example, 4 multi-scale images with original resolution, first low resolution, second low resolution and high resolution are respectively calculated as a 1 st scale image, a 2 nd scale image, a 3 rd scale image and a 4 th scale image, n is 2, and the license plate coarse positioning based on license plate features in a complex background is adopted to realize the license plate coarse positioning, Zhang Xin, Yuan Cheng, Guo ren Chun. "microcomputer information", 2006,22(33):306 and 308 ", so that the license plate coarse positioning is carried out on the i-th scale image, the license plate region coordinates are obtained, and the license plate region coordinates are put into a first license plate region queue tA [ i ].
The method for filtering the false license plate in the fourth step S4 may be an existing license plate filtering algorithm or method. For example, the license plate regions in the first license plate region queue are filtered using the relative density filtering and horizontal discrete point set measure filtering methods of the "a practical license plate location algorithm and implementation", Zhang jin Lin, screening Tree New, Tiger student, systems simulation bulletin, 2005,17(10): 2349-.
Further, the fifth step S5 includes:
a license plate fine positioning step S51, wherein the license plate fine positioning is carried out on all license plate areas in the first license plate area queue on the corresponding scale images, the average heights of a license plate fine positioning frame and license plate character blocks are obtained, the coordinates of the license plate fine positioning frame and the average heights of the license plate character blocks are restored to the original resolution, and then the license plate fine positioning frame and the license plate character blocks are placed into a second license plate area queue rectB;
an overlapping area searching step S52 of searching for an overlapping area of the license plate area in the second license plate area queue rectB, and placing the coordinates of the overlapping area and the average height of the character block into the third license plate area queue rectC;
a grouping calculation step S53, calculating the circumscribed rectangular frames RectT of all license plate areas in the third license plate area queue rectC; counting the upper limit UL of the license plate region identification range of the ith scale imageiAnd lower limit DLiIf the average height of the character block is located at λ × ULiAnd λ × DLiClassifying the license plate region into the ith scale image; counting the number of the classified license plate regions in each scale image;
a license plate fine positioning filtering step S54, if the number of the classified license plate regions in the ith scale image is more than 0, performing license plate fine positioning on the circumscribed rectangle frame RectT of the scale image, counting the average height of the character blocks of the fine positioning license plate region in the scale image, if the average height is located at lambda multiplied by ULiAnd λ × DLiOtherwise, filtering out the corresponding fine positioning license plate area;
and a fourth license plate area queue obtaining step S55, restoring the coordinates of the fine positioning license plate area of the scale image to the original resolution, and then placing the coordinates into a fourth license plate area queue rectD.
Further, the value range of the lambda is 0.3-0.2.
The accurate license plate positioning can be realized by the existing accurate license plate positioning algorithm or method.
In an embodiment, the third license plate area queue obtaining step S52 is: the method comprises the steps of calculating 4 multi-scale images with original resolution, first low resolution, second low resolution and high resolution, respectively calculating a 1 st scale image, a 2 nd scale image, a 3 rd scale image and a 4 th scale image, wherein n is 2, performing license plate fine positioning on each scale image of a second license plate area queue by adopting a method in a license plate fine positioning algorithm based on mathematical morphology, namely Mayongli, Xiauwa, microcomputer information 2008,24(1):200+233 and 234 document, obtaining the average height of a license plate fine positioning frame and license plate character blocks, reducing the coordinates of the license plate fine positioning frame and the average height of the license plate character blocks to the original resolution, and then placing the license plate fine positioning frame and the license plate character blocks into a third license plate area queue.
Further, the sixth step S6 includes:
a coincidence proportion calculation step S61 of respectively counting two arbitrary license plate regions PRect in the fourth license plate region queuepAnd selectqOverlap region select ofpqCalculating the coincidence region PRectpqRespectively in the license plate regionpAnd selectqOccupied area ratio SORpq/SRp、SORpq/SRqWherein SRp、SRq、SORpqRespectively a license plate region PRectpAnd selectqAnd a superposed region selectpqThe area of (d);
overlap Filter step S62 if SORpq/SRpAnd SORpq/SRqIf the number of the license plate is not less than the fourth threshold value, the license plate area PRect is reservedpAnd selectqFiltering out the license plate region with larger area from the license plate region with smaller area; if SORpq/SRpAnd SORpq/SRqIs not less than the fourth threshold value, the SOR is retainedpq/SRpAnd SORpq/SRqThe number plate area corresponding to the fourth threshold value is not less than the middle threshold value, and the SOR is filteredpq/SRpAnd SORpq/SRqThe number plate area is smaller than the number plate area corresponding to the fourth threshold value; if SORpq/SRpAnd SORpq/SRqAre all less than the fourth threshold value, the license plate area PRect is reserved at the same timepAnd selectq；
And a result output step S63 of outputting the license plate region retained in the fourth license plate region queue as a result.
Further, the value range of the fourth threshold is 0.7-0.9. Preferably, the fourth threshold is 0.8.
Fig. 2 is a frame diagram of a multi-scale positioning fusion device for a license plate according to the present invention. As shown in fig. 2, the multi-scale positioning fusion apparatus for license plates according to the present invention comprises:
a gray image input or acquisition module 1 for inputting or acquiring a gray image;
the multi-scale image acquisition module 2 is used for carrying out multi-scale processing on the gray level image to acquire a multi-scale image;
the license plate coarse positioning module 3 is used for performing license plate area coarse positioning on the multi-scale image to obtain a first license plate area queue;
the false license plate filtering module 4 is used for filtering the first license plate area queue according to a false license plate filtering method;
the license plate fine positioning and filtering module 5 is used for accurately positioning the first license plate area queue, restoring the coordinates to the original resolution, acquiring a second license plate area queue, searching an overlapping area of the second license plate area queue to acquire a third license plate area queue, and performing grouping fine positioning and filtering on the third license plate area queue to acquire a fourth license plate area queue;
and the license plate contact ratio filtering module 6 is used for filtering the contact ratio of the fourth license plate region queue to obtain a filtered license plate region and outputting the filtered license plate region.
The grayscale image input or acquisition module 1 may acquire a grayscale image through an image acquirer, or may directly input the grayscale image through an image storage device.
Further, the multi-scale image acquisition module 2 includes:
the scaling frequency calculation module 21 is configured to use the grayscale image as an original resolution image, and calculate the compression frequency and the scaling frequency according to a license plate width recognition range of the original resolution image and an expanded license plate width recognition range requirement;
a low-resolution image obtaining module 22, configured to extract or compress the original resolution image according to the compression frequency, so as to obtain a low-resolution image;
a high-resolution image obtaining module 23, configured to perform interpolation or amplification processing on the original high-resolution image according to the amplification times, so as to obtain a high-resolution image;
and a multi-scale image output module 24, configured to output the original resolution image, the low resolution image, and the high resolution image as multi-scale images.
Further, the scaling number calculation module 21 includes: the license plate width recognition range for the original resolution image is [ WR1,WR2]The requirement of the expanded license plate width recognition range is [ EWR1,EWR2]Calculating
Taking the integer as the amplification number, and calculating
And taking an integer as the compression time, and n is a scaling exponent.
Further, the value range of n is 2-4.
Further, the license plate fine positioning and filtering module 5 includes:
the license plate fine positioning module 51 is used for carrying out fine positioning on license plates on the images with the corresponding scales on all license plate areas in the first license plate area queue, acquiring the average heights of a license plate fine positioning frame and license plate character blocks, restoring the coordinates of the license plate fine positioning frame and the average heights of the license plate character blocks to the original resolution, and then placing the license plate fine positioning frame and the license plate character blocks into a second license plate area queue rectB;
an overlap area searching module 52, configured to search an overlap area of the license plate area in the second license plate area queue rectB, and place the coordinates of the overlap area and the average height of the character block into a third license plate area queue rectC;
the grouping calculation module 53 is configured to calculate circumscribed rectangular frames rect of all license plate regions in the third license plate region queue rectC; counting the upper limit UL of the license plate region identification range of the ith scale imageiAnd lower limit DLiIf the average height of the character block is located at λ × ULiAnd λ × DLiClassifying the license plate region into the ith scale image; counting the number of the classified license plate regions in each scale image;
a license plate fine positioning and filtering module 54 for determining if the ith ruler isIf the number of the classified license plate regions in the dimensional image is greater than 0, carrying out license plate fine positioning on the circumscribed rectangular frame RectT of the dimensional image, counting the average height of the character blocks of the fine positioning license plate regions in the dimensional image, and if the average height is located at lambda multiplied by ULiAnd λ × DLiOtherwise, filtering out the corresponding fine positioning license plate area;
and the fourth license plate area queue obtaining module 55 is configured to restore the coordinates of the fine positioning license plate area of the scale image to the original resolution, and then place the coordinates into a fourth license plate area queue rectD.
Further, the value range of the lambda is 0.3-0.2.
Further, the license plate contact ratio filtering module 6 includes:
a coincidence proportion calculating module 61 for respectively counting two arbitrary license plate regions select in the fourth license plate region queuepAnd selectqOverlap region select ofpqCalculating the coincidence region PRectpqRespectively in the license plate regionpAnd selectqOccupied area ratio SORpq/SRp、SORpq/SRqWherein SRp、SRq、SORpqRespectively a license plate region PRectpAnd selectqAnd a superposed region selectpqThe area of (d);
coincidence filtering module 62 for if SORpq/SRpAnd SORpq/SRqIf the number of the license plate is not less than the fourth threshold value, the license plate area PRect is reservedpAnd selectqFiltering out the license plate region with larger area from the license plate region with smaller area; if SORpq/SRpAnd SORpq/SRqIs not less than the fourth threshold value, the SOR is retainedpq/SRpAnd SORpq/SRqThe number plate area corresponding to the fourth threshold value is not less than the middle threshold value, and the SOR is filteredpq/SRpAnd SORpq/SRqThe number plate area is smaller than the number plate area corresponding to the fourth threshold value; if SORpq/SRpAnd SORpq/SRqAre all less than the fourth threshold value, the license plate area PRect is reserved at the same timepAnd PRectq；
And a result output module 63, configured to output the license plate area reserved in the fourth license plate area queue as a result.
Further, the value range of the fourth threshold is 0.7-0.9. Preferably, the fourth threshold is 0.8.
Compared with the existing license plate positioning technology, the multi-scale positioning fusion method and device for the license plate can quickly position the license plate and effectively improve the width recognition range of the license plate.
While the foregoing is directed to the preferred embodiment of the present invention, and is not intended to limit the scope of the invention, it will be understood that the invention is not limited to the embodiments described herein, which are described to assist those skilled in the art in practicing the invention. Further modifications and improvements may readily occur to those skilled in the art without departing from the spirit and scope of the invention, and it is intended that the invention be limited only by the terms and scope of the appended claims, as including all alternatives and equivalents which may be included within the spirit and scope of the invention as defined by the appended claims.
Claims (13)
1. A multi-scale positioning fusion method for a license plate is characterized by comprising the following steps:
the method comprises the steps of firstly, inputting or collecting a gray image;
the second step, carry on the multi-scale processing to the gray level picture, obtain the multi-scale picture;
thirdly, carrying out coarse positioning on the license plate area of the multi-scale image to obtain a first license plate area queue;
fourthly, filtering the first license plate area queue according to a false license plate filtering method;
a fifth step of accurately positioning the first license plate area queue and reducing the coordinates to the original resolution to obtain a second license plate area queue, searching the overlapping area of the second license plate area queue to obtain a third license plate area queue, and finely positioning and filtering the third license plate area queue in groups to obtain a fourth license plate area queue;
and sixthly, filtering the contact ratio of the fourth license plate area queue to obtain a filtered license plate area and outputting the filtered license plate area.
2. The method of claim 1, wherein the second step comprises:
a scaling frequency calculation step, namely taking the gray level image as an original resolution image, and respectively calculating the compression frequency and the amplification frequency according to the license plate width recognition range of the original resolution image and the expanded license plate width recognition range requirement;
a low resolution image acquisition step of extracting or compressing the original resolution image according to the compression times to acquire a low resolution image;
a high-resolution image acquisition step of performing interpolation or amplification processing on the original high-resolution image according to the amplification times to acquire a high-resolution image;
and a multi-scale image output step of outputting the original resolution image, the low resolution image and the high resolution image as multi-scale images.
3. The method of claim 2, wherein the scaling number calculation step comprises: the license plate width recognition range for the original resolution image is [ WR1,WR2]The requirement of the expanded license plate width recognition range is [ EWR1,EWR2]Calculating
Taking the integer as the amplification number, and calculating
And taking an integer as the compression time, and n is a scaling exponent.
4. The method of claim 3, wherein n has a value in the range of 2 to 4.
5. The method of claim 1, wherein the fifth step comprises:
a license plate fine positioning step, wherein the license plates of all license plate areas in the first license plate area queue are finely positioned on the corresponding scale image, the average heights of a license plate fine positioning frame and license plate character blocks are obtained, the coordinates of the license plate fine positioning frame and the average heights of the license plate character blocks are restored to the original resolution, and then the license plate fine positioning frame and the license plate character blocks are placed into a second license plate area queue rectB;
an overlapping area searching step, namely searching the overlapping area of the license plate area in the second license plate area queue rectB, and putting the coordinates of the overlapping area and the average height of the character block into a third license plate area queue rectC;
grouping calculation, namely calculating the circumscribed rectangular frames RectT of all license plate areas in the third license plate area queue rectC; counting the upper limit UL of the license plate region identification range of the ith scale imageiAnd lower limit DLiIf the average height of the character block is located at λ × ULiAnd λ × DLiClassifying the license plate region into the ith scale image; counting the number of the classified license plate regions in each scale image;
and (3) a license plate fine positioning filtering step, wherein if the number of the classified license plate regions in the ith scale image is greater than 0, the license plate fine positioning is carried out on the circumscribed rectangular frame RectT of the scale image, the average height of the character blocks of the finely positioned license plate region in the scale image is counted, and if the average height is located at lambda multiplied by ULiAnd λ × DLiOtherwise, filtering out the corresponding fine positioning license plate area;
and a fourth license plate area queue obtaining step, namely restoring the coordinates of the fine positioning license plate area of the scale image to the original resolution ratio, and then placing the coordinates into a fourth license plate area queue rectD.
6. The method of claim 5, wherein λ is in the range of 0.3-0.2.
7. The method of claim 1, wherein the sixth step comprises:
a coincidence proportion calculation step, namely respectively counting two random license plate regions PRect in the fourth license plate region queuepAnd selectqOverlap region select ofpqCalculating the coincidence region PRectpqRespectively in the license plate regionpAnd selectqOccupied area ratio SORpq/SRp、SORpq/SRqWherein SRp、SRq、SORpqRespectively a license plate region PRectpAnd selectqAnd a superposed region selectpqThe area of (d);
coincidence of filtering step if SORpq/SRpAnd SORpq/SRqIf the number of the license plate is not less than the fourth threshold value, the license plate area PRect is reservedpAnd selectqFiltering out the license plate region with larger area from the license plate region with smaller area; if SORpq/SRpAnd SORpq/SRqIs not less than the fourth threshold value, the SOR is retainedpq/SRpAnd SORpq/SRqThe number plate area corresponding to the fourth threshold value is not less than the middle threshold value, and the SOR is filteredpq/SRpAnd SORpq/SRqThe number plate area is smaller than the number plate area corresponding to the fourth threshold value; if SORpq/SRpAnd SORpq/SRqAre all less than the fourth threshold value, the license plate area PRect is reserved at the same timepAnd selectq；
And a result output step, namely taking the license plate area reserved in the fourth license plate area queue as a result and outputting the result.
8. The method of claim 7, wherein the fourth threshold value ranges from 0.7 to 0.9.
9. A multiscale positioning fusion device for a license plate, the device comprising:
the gray level image input or acquisition module is used for inputting or acquiring a gray level image;
the multi-scale image acquisition module is used for carrying out multi-scale processing on the gray level image to acquire a multi-scale image;
the license plate coarse positioning module is used for performing license plate area coarse positioning on the multi-scale image to obtain a first license plate area queue;
the false license plate filtering module is used for filtering the first license plate area queue according to a false license plate filtering method;
the license plate fine positioning and filtering module is used for accurately positioning the first license plate area queue, restoring the coordinates to the original resolution, acquiring a second license plate area queue, searching an overlapping area of the second license plate area queue to acquire a third license plate area queue, and performing grouping fine positioning and filtering on the third license plate area queue to acquire a fourth license plate area queue;
and the license plate contact ratio filtering module is used for filtering the contact ratio of the fourth license plate region queue to obtain a filtered license plate region and outputting the filtered license plate region.
10. The apparatus of claim 9, wherein the multi-scale image acquisition module comprises:
the scaling frequency calculation module is used for taking the gray level image as an original resolution image, and respectively calculating the compression frequency and the amplification frequency according to the license plate width recognition range of the original resolution image and the expanded license plate width recognition range requirement;
the low-resolution image acquisition module is used for extracting or compressing the original resolution image according to the compression times to acquire a low-resolution image;
the high-resolution image acquisition module is used for carrying out interpolation or amplification processing on the original high-resolution image according to the amplification times to acquire a high-resolution image;
and the multi-scale image output module is used for outputting the original resolution image, the low resolution image and the high resolution image as multi-scale images.
11. The apparatus of claim 10, wherein the scaling times calculation module comprises: the license plate width recognition range for the original resolution image is [ WR1，WR2]The requirement of the expanded license plate width recognition range is [ EWR1，EWR2]Calculating
Taking the integer as the amplification number, and calculating
And taking an integer as the compression time, and n is a scaling exponent.
12. The apparatus of claim 9, wherein the license plate fine positioning and filtering module comprises:
the license plate fine positioning module is used for carrying out fine positioning on license plates on the images with the corresponding scales on all license plate areas in the first license plate area queue to obtain the average heights of a license plate fine positioning frame and license plate character blocks, reducing the coordinates of the license plate fine positioning frame and the average heights of the license plate character blocks to the original resolution, and then placing the license plate fine positioning frame and the license plate character blocks into a second license plate area queue rectB;
the overlapping area searching module is used for searching the overlapping area of the license plate area in the second license plate area queue rectB and putting the coordinates of the overlapping area and the average height of the character block into a third license plate area queue rectC;
the grouping calculation module is used for calculating the circumscribed rectangular frames RectT of all license plate areas in the third license plate area queue rectC; counting the upper limit UL of the license plate region identification range of the ith scale imageiAnd lower limit DLiIf the average height of the character block is located at λ × ULiAnd λ × DLiClassifying the license plate region into the ith scale image; counting the number of the classified license plate regions in each scale image;
a license plate fine positioning filtering module, configured to perform license plate fine positioning on an external rectangular frame RectT of the scale image if the number of classified license plate regions in the ith scale image is greater than 0, count an average height of character blocks in the fine positioning license plate region in the scale image, and if the average height is located at λ × ULiAnd λ × DLiBetweenReserving the license plate area, otherwise, filtering the corresponding fine positioning license plate area;
and the fourth license plate area queue acquisition module is used for restoring the coordinates of the fine positioning license plate area of the scale image to the original resolution ratio and then placing the coordinates into a fourth license plate area queue rectD.
13. The apparatus of claim 9, wherein the license plate overlap ratio filter module comprises:
a coincidence proportion calculation module for respectively counting two arbitrary license plate regions PRect in the fourth license plate region queuepAnd selectqOverlap region select ofpqCalculating the coincidence region PRectpqRespectively in the license plate regionpAnd selectqOccupied area ratio SORpq/SRp、SORpq/SRqWherein SRp、SRq、SORpqRespectively a license plate region PRectpAnd selectqAnd a superposed region selectpqThe area of (d);
coincidence filter module for if SORpq/SRpAnd SORpq/SRqIf the number of the license plate is not less than the fourth threshold value, the license plate area PRect is reservedpAnd selectqFiltering out the license plate region with larger area from the license plate region with smaller area; if SORpq/SRpAnd SORpq/SRqIs not less than the fourth threshold value, the SOR is retainedpq/SRpAnd SORpq/SRqThe number plate area corresponding to the fourth threshold value is not less than the middle threshold value, and the SOR is filteredpq/SRpAnd SORpq/SRqThe number plate area is smaller than the number plate area corresponding to the fourth threshold value; if SORpq/SRpAnd SORpq/SRqAre all less than the fourth threshold value, the license plate area PRect is reserved at the same timepAnd selectq；
And the result output module is used for outputting the license plate area reserved in the fourth license plate area queue as a result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810582289.0A CN108776792B (en) | 2018-06-07 | 2018-06-07 | Multi-scale positioning fusion method and device for license plate |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810582289.0A CN108776792B (en) | 2018-06-07 | 2018-06-07 | Multi-scale positioning fusion method and device for license plate |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108776792A CN108776792A (en) | 2018-11-09 |
| CN108776792B true CN108776792B (en) | 2021-09-17 |
Family
ID=64025661
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810582289.0A Active CN108776792B (en) | 2018-06-07 | 2018-06-07 | Multi-scale positioning fusion method and device for license plate |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108776792B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0800146A2 (en) * | 1996-03-26 | 1997-10-08 | Sharp Kabushiki Kaisha | Character recognition method |
| CN103699905A (en) * | 2013-12-27 | 2014-04-02 | 深圳市捷顺科技实业股份有限公司 | Method and device for positioning license plate |
| CN103793708A (en) * | 2014-03-05 | 2014-05-14 | 武汉大学 | Multi-scale license plate precise locating method based on affine correction |
| CN106407959A (en) * | 2016-11-07 | 2017-02-15 | 湖南源信光电科技有限公司 | Low-illumination complicated background license plate positioning method based on wavelet transform and SVM |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8798325B2 (en) * | 2012-02-21 | 2014-08-05 | Xerox Corporation | Efficient and fault tolerant license plate matching method |
-
2018
- 2018-06-07 CN CN201810582289.0A patent/CN108776792B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0800146A2 (en) * | 1996-03-26 | 1997-10-08 | Sharp Kabushiki Kaisha | Character recognition method |
| CN103699905A (en) * | 2013-12-27 | 2014-04-02 | 深圳市捷顺科技实业股份有限公司 | Method and device for positioning license plate |
| CN103793708A (en) * | 2014-03-05 | 2014-05-14 | 武汉大学 | Multi-scale license plate precise locating method based on affine correction |
| CN106407959A (en) * | 2016-11-07 | 2017-02-15 | 湖南源信光电科技有限公司 | Low-illumination complicated background license plate positioning method based on wavelet transform and SVM |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108776792A (en) | 2018-11-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108986465B (en) | Method, system and terminal equipment for detecting traffic flow | |
| CN104700099B (en) | The method and apparatus for recognizing traffic sign | |
| CN104239867B (en) | License plate locating method and system | |
| CN109583483B (en) | Target detection method and system based on convolutional neural network | |
| CN112733822B (en) | End-to-end text detection and identification method | |
| CN103577817B (en) | Form recognition method and apparatus | |
| CN102163284B (en) | Chinese environment-oriented complex scene text positioning method | |
| CN106971185B (en) | License plate positioning method and device based on full convolution network | |
| CN105426891B (en) | Registration number character dividing method and its system based on image | |
| CN112329776B (en) | License plate detection method and device based on improved CenterNet network | |
| CN111582339B (en) | Vehicle detection and recognition method based on deep learning | |
| CN109583442B (en) | False license plate detection method and device based on line segment detection | |
| CN103971097A (en) | Vehicle license plate recognition method and system based on multiscale stroke models | |
| CN107480585A (en) | Object detection method based on DPM algorithms | |
| CN110490150A (en) | A kind of automatic auditing system of picture violating the regulations and method based on vehicle retrieval | |
| CN115761563A (en) | River surface flow velocity calculation method and system based on optical flow measurement and calculation | |
| CN105608689A (en) | Method and device for eliminating image feature mismatching for panoramic stitching | |
| CN116052105A (en) | Pavement crack identification classification and area calculation method, system, equipment and terminal | |
| CN116704490B (en) | License plate recognition method, license plate recognition device and computer equipment | |
| CN108776792B (en) | Multi-scale positioning fusion method and device for license plate | |
| CN111488839B (en) | Target detection method and target detection system | |
| CN112446353A (en) | Video image trace line detection method based on deep convolutional neural network | |
| CN117456441A (en) | Monitoring method and system for rust area expansion by combining change area identification | |
| CN113011293B (en) | Real-time extraction method for lane line parameters | |
| CN112950954B (en) | Intelligent parking license plate recognition method based on high-position camera |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |