CN110533748A - Seal minimizing technology and equipment - Google Patents
Seal minimizing technology and equipment Download PDFInfo
- Publication number
- CN110533748A CN110533748A CN201910802744.8A CN201910802744A CN110533748A CN 110533748 A CN110533748 A CN 110533748A CN 201910802744 A CN201910802744 A CN 201910802744A CN 110533748 A CN110533748 A CN 110533748A
- Authority
- CN
- China
- Prior art keywords
- seal
- image
- text image
- minimum outsourcing
- outsourcing rectangle
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/30—Authentication, i.e. establishing the identity or authorisation of security principals
- G06F21/31—User authentication
- G06F21/36—User authentication by graphic or iconic representation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T11/00—2D [Two Dimensional] image generation
- G06T11/60—Editing figures and text; Combining figures or text
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR 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; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20021—Dividing image into blocks, subimages or windows
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20112—Image segmentation details
- G06T2207/20132—Image cropping
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Security & Cryptography (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computer Hardware Design (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The object of the present invention is to provide a kind of seal minimizing technology and equipment, the present invention can solve existing tradition and go seal method to illumination, seal color, the problem of bad adaptability of the seal depth, the present invention can be used minimum outsourcing rectangle and detect car test table seal, it is cut again by seal and the seal minimizing technology based on image reconstruction is removed seal and retains the text information under seal, image block after cutting is finally spliced into the text image after removal seal, text image after the removal seal spliced is replaced into the seal region in the minimum outsourcing rectangle in the original text image.Red chapter in car test table can be removed and retain the text image covered by red chapter by the present invention, it does homework for the processing of subsequent car test table character, what the invention was adapted to all car test tables goes seal demand, its is adaptable, maintenance is simple, applied value is high, can reduce Project-developing cost.
Description
Technical field
The present invention relates to computer field more particularly to a kind of seal minimizing technology and equipment.
Background technique
During Vehicle Security examines (hereinafter referred to as: car test), the detection and identification of car test table class occupy
Very big specific gravity, and it is to lead to the important original that critical field detection efficiency is low or recognition efficiency is low in critical field that seal, which is covered,
One of because.
It is using traditional image processing method, to illumination, seal color, the seal depth that existing table, which removes seal all,
Adaptability it is poor, seal generally can not be removed well clean, or the character under seal can be removed together, after influencing
Continuous detection and identification.
Summary of the invention
It is an object of the present invention to provide a kind of seal minimizing technology and equipment.
According to an aspect of the invention, there is provided a kind of seal minimizing technology, this method comprises:
The original text image for obtaining seal to be removed, the minimum for obtaining the seal region in the original text image are outer
Packet rectangle；
Seal text image in the minimum outsourcing rectangle is cut to multiple images block；To the print in each image block
Chapter is removed to obtain the plain text image of corresponding seal bottom；By correspondence image block in the minimum outsourcing rectangle
Each plain text image mosaic is the text image after removing seal by home position；
Text image after the removal seal spliced is replaced in the minimum outsourcing rectangle in the original text image
Seal region.
Further, in the above method, the minimum outsourcing rectangle in the seal region in the original text image is obtained, is wrapped
It includes:
By the first object detection model of SSD structure, the minimum in the seal region in the original text image is obtained
(x, y, w, h) value of outsourcing rectangle, wherein wherein x, y respectively indicate the upper left corner of the minimum outsourcing rectangle abscissa and
Ordinate, w, h respectively indicate the width and height of the minimum outsourcing rectangle.
Further, in the above method, by first object detection model, the seal in the original text image is obtained
Before (x, y, w, h) value of the minimum outsourcing rectangle in region, further includes:
It is driven and is trained by following loss function L (m, c, l, g), obtain corresponding first object detection model:
L (m, c, l, g)=σ * Lconf(x, c)+β * Lloc(m, l, g)
Wherein, for σ presentation class loss function to the contribution proportion of whole loss function, β expression position returns loss function
To the contribution proportion of whole loss function, first object detection model predicts a certain position to be seal and probability when being p, m's
Value is 1, and m value is that 0, c indicates that the value of m is 1 and the correct probability of model prediction when front frame is target in the case of remaining, l the
The position for the seal that one target detection model prediction is arrived, g are the practical position of seal, and Pos indicates that first object detects mould
All possible position where the seal that type predicts, i indicate that the index of seal position in Pos, ij indicate first object inspection
J-th of seal that model prediction corresponds to the upper practical position of seal to i-th of seal is surveyed, k indicates the location information of seal,
The place that Neg indicates that first object detection model detects is not seal.
Further, in the above method, the seal text image in the minimum outsourcing rectangle is cut to multiple images
Block, comprising:
Seal text image in the minimum outsourcing rectangle having a size of (w, h), which is cut to multiple sizes, is respectively
(m, m) image block, m are positive number, wherein
If w is divided exactly by m, and h is divided exactly by m, then will be having a size of the seal text diagram in the minimum outsourcing rectangle of (w, h)
As being cut to wh/m2A image block；
If w is not divided exactly by m, and h is divided exactly by m, then will be having a size of the seal text in the minimum outsourcing rectangle of (w, h)
Image cropping is (w-w%m) h/m2A image block；
If w is divided exactly by m, and h is not divided exactly by m, then will be having a size of the seal text in the minimum outsourcing rectangle of (w, h)
Image cropping is into (h-h%m) w/m2A image block；
If w is not divided exactly by m, and h is not divided exactly by m, then will be having a size of the seal text in the minimum outsourcing rectangle of (w, h)
This image cropping is (h-h%m) (w-w%m)/m2A image block.
Further, in the above method, the seal in each image block is removed to obtain corresponding seal bottom
Plain text image, including
By the second target detection model based on Super-resolution reconstruction established model IDN, to the seal in each image block into
Row is gone divided by obtaining the plain text image of corresponding seal bottom.
Further, right by the second target detection model based on Super-resolution reconstruction established model IDN in the above method
Before seal in each image block is removed to obtain the plain text image of corresponding seal bottom, further includes:
Pass through loss function LMAEDriving training obtains the second target inspection based on Super-resolution reconstruction established model IDN accordingly
Survey model:
Wherein, N=m*m*3, IiFor corresponding pixel value at the point i in current image to be reconstructed, Ii' it is in label image
Point i at corresponding pixel value, current image to be reconstructed is the image that the second target detection model prediction goes out, the label image
The effect image that should be actually reconstructed into for the current image to be reconstructed.
Further, in the above method, home position of the correspondence image block in the minimum outsourcing rectangle will be each pure
Splicing document images be remove seal after text image in,
If two adjacent image blocks have overlapping in the minimum outsourcing rectangle, weight described in the text image after removing seal
The pixel value of folded part, with the corresponding plain text image of adjacent image block that sorts rearward in the minimum outsourcing rectangle
Subject to pixel value.
According to another aspect of the present invention, a kind of seal eliminating equipment is also provided, which includes:
Seal region acquisition device obtains the urtext figure for obtaining the original text image of seal to be removed
The minimum outsourcing rectangle in the seal region as in；
Seal cuts removal splicing apparatus, multiple for the seal text image in the minimum outsourcing rectangle to be cut to
Image block；Seal in each image block is removed to obtain the plain text image of corresponding seal bottom；By corresponding diagram
It is the text diagram after removing seal by each plain text image mosaic as home position of the block in the minimum outsourcing rectangle
Picture；
It is outer to be replaced the minimum in the original text image by alternative for text image after the removal seal spliced
Seal region in packet rectangle.
According to another aspect of the present invention, a kind of computer-readable medium is also provided, computer-readable finger is stored thereon with
It enables, the computer-readable instruction can be executed by processor to realize method described in any of the above embodiments.
According to another aspect of the present invention, a kind of equipment for handling in network equipment client information, the equipment are also provided
Including the memory for storing computer program instructions and the processor for executing program instructions, which is characterized in that when this
When computer program instructions are executed by the processor, triggers the equipment and execute method described in any of the above embodiments.
Compared with prior art, the present invention can solve existing tradition and go seal method deep to illumination, seal color, seal
The problem of shallow bad adaptability, the present invention can be used minimum outsourcing rectangle and detect car test table seal, then pass through seal
It cuts and the seal minimizing technology based on image reconstruction is removed seal and retains the text information under seal, will finally cut out
Image block after cutting is spliced into the text image after removal seal, will be described in the text image replacement after the removal seal that splice
The seal region in minimum outsourcing rectangle in original text image.Red chapter in car test table can be removed and be protected by the present invention
The text image covered by red chapter is stayed, is done homework for the processing of subsequent car test table character, which is adapted to all
Car test table goes seal demand, and adaptable, maintenance is simple, and applied value is high, can reduce Project-developing cost.
Present invention can apply to going in seal application for car test table, provide for the detection and identification of subsequent car test table
Precondition is compared with existing method, and the present invention is adapted to all car test table seals, and to illumination, seal type and
The depth is insensitive, applied widely, greatly reduces and does the manpower expended when specially treated and computing resource for every kind of seal,
Cost is reduced, it is easy to maintain.
Detailed description of the invention
By reading a detailed description of non-restrictive embodiments in the light of the attached drawings below, of the invention other
Feature, objects and advantages will become more apparent upon:
Fig. 1 shows the schematic diagram of the seal minimizing technology of one embodiment of the invention；
Fig. 2 shows the flow charts of the seal minimizing technology of one embodiment of the invention；
Fig. 3 shows the seal cutting schematic diagram of one embodiment of the invention；
What Fig. 4 showed one embodiment of the invention goes seal network diagram；
What Fig. 5 showed one embodiment of the invention removes seal network details schematic diagram；
Fig. 6 shows the algorithm flow chart of the seal minimizing technology of one embodiment of the invention.
The same or similar appended drawing reference represents the same or similar component in attached drawing.
Specific embodiment
Present invention is further described in detail with reference to the accompanying drawing.
In a typical configuration of this application, terminal, the equipment of service network and trusted party include one or more
Processor (CPU), input/output interface, network interface and memory.
Memory may include the non-volatile memory in computer-readable medium, random access memory (RAM) and/or
The forms such as Nonvolatile memory, such as read-only memory (ROM) or flash memory (flash RAM).Memory is computer-readable medium
Example.
Computer-readable medium includes permanent and non-permanent, removable and non-removable media can be by any method
Or technology come realize information store.Information can be computer readable instructions, data structure, the module of program or other data.
The example of the storage medium of computer includes, but are not limited to phase change memory (PRAM), static random access memory (SRAM), moves
State random access memory (DRAM), other kinds of random access memory (RAM), read-only memory (ROM), electric erasable
Programmable read only memory (EEPROM), flash memory or other memory techniques, read-only disc read only memory (CD-ROM) (CD-ROM),
Digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or
Any other non-transmission medium, can be used for storage can be accessed by a computing device information.As defined in this article, computer
Readable medium does not include non-temporary computer readable media (transitory media), such as the data-signal and carrier wave of modulation.
As illustrated in fig. 1 and 2, the present invention provides a kind of seal minimizing technology, which comprises
Step S1 obtains the original text image of seal to be removed；
Step S2 obtains the minimum outsourcing rectangle in the seal region in the original text image；
Seal text image in the minimum outsourcing rectangle is cut to multiple images block by step S3；To each image
Seal in block is removed to obtain the plain text image of corresponding seal bottom；By correspondence image block in the minimum outsourcing
Each plain text image mosaic is the text image after removing seal by the home position in rectangle；
Here, according to the above minimum outsourcing rectangle seal text image in seal region in car test table can be obtained, so
It is cut afterwards by seal, the text information under seal removal reservation seal and seal splice, the text diagram after obtaining removal seal
Picture；
Text image after the removal seal spliced is replaced the minimum outsourcing in the original text image by step S4
Seal region in rectangle.
Here, finally piecing back the text image after the removal seal spliced in original text image together, result is obtained
Image, and result images are exported, in case subsequent use.
The present invention can solve existing tradition and go seal method to illumination, seal color, the bad adaptability of the seal depth
Problem, the present invention can be used minimum outsourcing rectangle and detect car test table seal, then cut and be based on image by seal
The seal minimizing technology of reconstruction is removed seal and retains the text information under seal, finally spells the image block after cutting
Text image after being connected in removal seal replaces the text image after the removal seal spliced in the original text image
Minimum outsourcing rectangle in seal region.Red chapter in car test table can be removed and be retained and covered by red chapter by the present invention
Text image does homework for the processing of subsequent car test table character, which is adapted to all car test table and goes to print
Chapter demand, adaptable, maintenance is simple, and applied value is high, can reduce Project-developing cost.
Present invention can apply to going in seal application for car test table, provide for the detection and identification of subsequent car test table
Precondition is compared with existing method, and the present invention is adapted to all car test table seals, and to illumination, seal type and
The depth is insensitive, applied widely, greatly reduces and does the manpower expended when specially treated and computing resource for every kind of seal,
Cost is reduced, it is easy to maintain.
In one embodiment of seal minimizing technology of the invention, step S2 obtains the seal area in the original text image
The minimum outsourcing rectangle in domain, comprising:
By the first object detection model of SSD structure, the minimum in the seal region in the original text image is obtained
(x, y, w, h) value of outsourcing rectangle, wherein wherein x, y respectively indicate the upper left corner of the minimum outsourcing rectangle abscissa and
Ordinate, w, h respectively indicate the width and height of the minimum outsourcing rectangle.
Here, first object detection model is derived from SSD structure, output result includes that information has (label, x, y, w, h),
In, label represents the classification of content in minimum outsourcing rectangle, for example, classification is seal region, (x, y, w, h) four value difference
The top left co-ordinate and its width and height of the minimum outsourcing rectangle of expression.
The present invention uses the model structure of SSD (the more frame detectors of single-shot), and returning (x, y, w, h) four values indicates seal area
Domain, with the acquisition seal region of high efficient and reliable.
In one embodiment of seal minimizing technology of the invention, by first object detection model, the urtext is obtained
Before (x, y, w, h) value of the minimum outsourcing rectangle in the seal region in image, further includes:
It is driven and is trained by following loss function L (m, c, l, g), obtain corresponding first object detection model:
L (m, c, l, g)=σ * Lconf(x, c)+β * Lloc(m, l, g)
Wherein, for σ presentation class loss function to the contribution proportion of whole loss function, β expression position returns loss function
To the contribution proportion of whole loss function, first object detection model predicts a certain position to be seal and probability when being p, m's
Value is 1, and m value is that 0, c indicates that the value of m is 1 and the correct probability of model prediction when front frame is target in the case of remaining, l the
The position for the seal that one target detection model prediction is arrived, g are the practical position of seal, and Pos indicates that first object detects mould
All possible position where the seal that type predicts, i indicate that the index of seal position in Pos, ij indicate first object inspection
J-th of seal that model prediction corresponds to the upper practical position of seal to i-th of seal is surveyed, k indicates the location information of seal,
The place that Neg indicates that first object detection model detects is not seal.
Here, the car test form image for meeting real distribution can be first obtained when training first object detection model,
Then table seal is labeled with minimum outsourcing rectangle；Finally, learning rate 0.03 drives according to loss function L (m, c, l, g)
Movable model training, obtains the first object detection model.
In one embodiment of seal minimizing technology of the invention, in step S3, by the seal text in the minimum outsourcing rectangle
This image cropping is multiple images block, comprising:
Seal text image in the minimum outsourcing rectangle having a size of (w, h), which is cut to multiple sizes, is respectively
(m, m) image block, m are positive number, wherein
If w is divided exactly by m, and h is divided exactly by m, then will be having a size of the seal text diagram in the minimum outsourcing rectangle of (w, h)
As being cut to wh/m2A image block；
If w is not divided exactly by m, and h is divided exactly by m, then will be having a size of the seal text in the minimum outsourcing rectangle of (w, h)
Image cropping is (w-w%m) h/m2A image block；
If w is divided exactly by m, and h is not divided exactly by m, then will be having a size of the seal text in the minimum outsourcing rectangle of (w, h)
Image cropping is into (h-h%m) w/m2A image block；
If w is not divided exactly by m, and h is not divided exactly by m, then will be having a size of the seal text in the minimum outsourcing rectangle of (w, h)
This image cropping is (h-h%m) (w-w%m)/m2A image block.
Here, since hardware computing capability is restricted, and the original seal size detected is different, therefore it is removed in seal
Before need to cut seal image so that its is fixed-size while adapting to hardware.Assuming that the seal ruler that detection obtains
Very little is (w, h), and the size for needing to be cut into is (m, m), can obtain each image block according to Pruning strategy as above, specific to cut out
Cutting strategy can be found in Fig. 3.
The tile size (m, m) for needing to be cut into, then, foundation can be set according to project demands and hardware condition
The present embodiment seal Pruning strategy ultimately produces red chapter block combination by red chapter image cropping at the hot chapter block of specified size.
In one embodiment of seal minimizing technology of the invention, in step S3, the seal in each image block is removed
To obtain the plain text image of corresponding seal bottom, including
By being based on Super-resolution reconstruction established model IDN (Fast and Accurate Single Image Super-
Resolution via Information Distillation Network, the quick single image oversubscription based on information distillation
Resolution network) the second target detection model, the seal in each image block is removed to obtain corresponding seal bottom
Plain text image.
Here, the present invention goes seal to use a kind of seal minimizing technology based on image reconstruction, this method is based on super
Resolution reconstruction model IDN, according to the present invention without being amplified to seal picture, therefore by the deconvolution amplifier section of master mould
It deletes, finally obtained model structure can be as shown in Figures 4 and 5.Wherein, Fig. 4 is to go seal model structure, and Fig. 5 is model knot
The details of enhancing module and compression module in structure.
In one embodiment of seal minimizing technology of the invention, pass through the second target based on Super-resolution reconstruction established model IDN
Detection model, before being removed to the seal in each image block to obtain the plain text image of corresponding seal bottom, also
Include:
Pass through loss function LMAEDriving training obtains the second target inspection based on Super-resolution reconstruction established model IDN accordingly
Survey model:
Wherein, N=m*m*3, IiFor corresponding pixel value at the point i in current image to be reconstructed, Ii' it is in label image
Point i at corresponding pixel value, current image to be reconstructed is the image that the second target detection model prediction goes out, the label image
The effect image that should be actually reconstructed into for the current image to be reconstructed.
Here, the image block execution that seal can obtain cutting after cutting goes seal to operate.
When the second target detection model of training, can first it obtain containing the form image of seal and its correspondence completely without print
The form image of chapter；
Then, image is obtained in table at red chapter as sample and clean table corresponding position image conduct
GroundTruth (learning objective)；
Then, red chapter picture and GroundTruth picture are cut respectively according to Pruning strategy above-mentioned；
According to loss function LMAE, the training of 0.01 driving model of learning rate, the second target detection model of acquisition.
In one embodiment of seal minimizing technology of the invention, in step S3, correspondence image block is in the minimum outsourcing rectangle
In home position, by each plain text image mosaic be remove seal after text image in,
If two adjacent image blocks have overlapping in the minimum outsourcing rectangle, weight described in the text image after removing seal
The pixel value of folded part, with the corresponding plain text image of adjacent image block that sorts rearward in the minimum outsourcing rectangle
Subject to pixel value.
Here, again can will according to the Pruning strategy of front after having executed seal operation to each image block
Went the image block of seal according to from left to right, sequence from top to bottom is successively spliced, if encountering image block has overlapping
Situation, is subject to the pixel value of the sequence corresponding plain text image of image block rearward.
Specifically, as shown in fig. 6, in one embodiment of seal minimizing technology of the invention,
Firstly, obtaining car test form image to be detected, the minimum outsourcing rectangle in seal region is obtained,
If minimum outsourcing rectangle obtains failure, directly exit；
If obtaining successfully, cut through seal, seal removal and seal splicing, realization removal table seal retain seal
Lower character information；
The Output of for ms of seal will be finally removed, in case subsequent use.
According to another aspect of the present invention, a kind of seal eliminating equipment is also provided, which is characterized in that the equipment includes:
Seal region acquisition device obtains the urtext figure for obtaining the original text image of seal to be removed
The minimum outsourcing rectangle in the seal region as in；
Seal cuts removal splicing apparatus, multiple for the seal text image in the minimum outsourcing rectangle to be cut to
Image block；Seal in each image block is removed to obtain the plain text image of corresponding seal bottom；By corresponding diagram
It is the text diagram after removing seal by each plain text image mosaic as home position of the block in the minimum outsourcing rectangle
Picture；
It is outer to be replaced the minimum in the original text image by alternative for text image after the removal seal spliced
Seal region in packet rectangle.
According to another aspect of the present invention, a kind of computer-readable medium is also provided, computer-readable finger is stored thereon with
It enables, the computer-readable instruction can be executed by processor to realize method described in any of the above embodiments.
According to another aspect of the present invention, a kind of equipment for handling in network equipment client information, the equipment are also provided
Including the memory for storing computer program instructions and the processor for executing program instructions, when the computer program refers to
When enabling by processor execution, triggers the equipment and execute method described in any of the above embodiments.
The detailed content of each equipment and storage medium embodiment of the invention, for details, reference can be made to the correspondences of each method embodiment
Part, here, repeating no more.
Obviously, those skilled in the art can carry out various modification and variations without departing from the essence of the application to the application
Mind and range.In this way, if these modifications and variations of the application belong to the range of the claim of this application and its equivalent technologies
Within, then the application is also intended to include these modifications and variations.
It should be noted that the present invention can be carried out in the assembly of software and/or software and hardware, for example, can adopt
With specific integrated circuit (ASIC), general purpose computer or any other realized similar to hardware device.In one embodiment
In, software program of the invention can be executed to implement the above steps or functions by processor.Similarly, of the invention
Software program (including relevant data structure) can be stored in computer readable recording medium, for example, RAM memory,
Magnetic or optical driver or floppy disc and similar devices.In addition, some of the steps or functions of the present invention may be implemented in hardware, example
Such as, as the circuit cooperated with processor thereby executing each step or function.
In addition, a part of the invention can be applied to computer program product, such as computer program instructions, when its quilt
When computer executes, by the operation of the computer, it can call or provide according to the method for the present invention and/or technical solution.
And the program instruction of method of the invention is called, it is possibly stored in fixed or moveable recording medium, and/or pass through
Broadcast or the data flow in other signal-bearing mediums and transmitted, and/or be stored according to described program instruction operation
In the working storage of computer equipment.Here, according to one embodiment of present invention including a device, which includes using
Memory in storage computer program instructions and processor for executing program instructions, wherein when the computer program refers to
When enabling by processor execution, method and/or skill of the device operation based on aforementioned multiple embodiments according to the present invention are triggered
Art scheme.
It is obvious to a person skilled in the art that invention is not limited to the details of the above exemplary embodiments, Er Qie
In the case where without departing substantially from spirit or essential attributes of the invention, the present invention can be realized in other specific forms.Therefore, no matter
From the point of view of which point, the present embodiments are to be considered as illustrative and not restrictive, and the scope of the present invention is by appended power
Benefit requires rather than above description limits, it is intended that all by what is fallen within the meaning and scope of the equivalent elements of the claims
Variation is included in the present invention.Any reference signs in the claims should not be construed as limiting the involved claims.This
Outside, it is clear that one word of " comprising " does not exclude other units or steps, and odd number is not excluded for plural number.That states in device claim is multiple
Unit or device can also be implemented through software or hardware by a unit or device.The first, the second equal words are used to table
Show title, and does not indicate any particular order.
Claims (10)
1. a kind of seal minimizing technology, which is characterized in that this method comprises:
The original text image for obtaining seal to be removed obtains the minimum outsourcing square in the seal region in the original text image
Shape；
Seal text image in the minimum outsourcing rectangle is cut to multiple images block；To the seal in each image block into
Row is gone divided by obtaining the plain text image of corresponding seal bottom；It is original in the minimum outsourcing rectangle by correspondence image block
Each plain text image mosaic is the text image after removing seal by position；
Text image after the removal seal spliced is replaced into the print in the minimum outsourcing rectangle in the original text image
Chapter region.
2. the method according to claim 1, wherein obtaining seal region in the original text image most
Small outsourcing rectangle, comprising:
By the first object detection model of SSD structure, the minimum outsourcing in the seal region in the original text image is obtained
(x, y, w, h) value of rectangle, wherein wherein x, y respectively indicate the abscissa and vertical seat in the upper left corner of the minimum outsourcing rectangle
Mark, w, h respectively indicate the width and height of the minimum outsourcing rectangle.
3. according to the method described in claim 2, it is characterized in that, obtaining the original text by first object detection model
Before (x, y, w, h) value of the minimum outsourcing rectangle in the seal region in this image, further includes:
It is driven and is trained by following loss function L (m, c, l, g), obtain corresponding first object detection model:
L (m, c, l, g)=σ * Lconf(x, c)+β * Lloc(m, l, g)
Wherein, for σ presentation class loss function to the contribution proportion of whole loss function, β expression position returns loss function to whole
The contribution proportion of bulk diffusion function, first object detection model predicts a certain position to be seal and probability when being p, and the value of m is
1, m value is that 0, c indicates that the value of m is 1 and the correct probability of model prediction when front frame is target in the case of remaining, and l is the first mesh
The position for the seal that mark detection model predicts, g are the practical position of seal, and Pos indicates that first object detection model is pre-
All possible position where the seal measured, i indicate that the index of seal position in Pos, ij indicate that first object detects mould
Type predicts j-th of seal of the corresponding upper practical position of seal of i-th of seal, and k indicates the location information of seal, Neg table
The place for showing that first object detection model detects is not seal.
4. the method according to claim 1, wherein the seal text image in the minimum outsourcing rectangle is cut out
It is cut to multiple images block, comprising:
It is respectively (m, m) that seal text image in the minimum outsourcing rectangle having a size of (w, h), which is cut to multiple sizes,
Image block, m are positive number, wherein
If w is divided exactly by m, and h is divided exactly by m, then will be cut out having a size of the seal text image in the minimum outsourcing rectangle of (w, h)
It is cut to wh/m2A image block；
If w is not divided exactly by m, and h is divided exactly by m, then will be having a size of the seal text image in the minimum outsourcing rectangle of (w, h)
It is cut to (w-w%m) h/m2A image block；
If w is divided exactly by m, and h is not divided exactly by m, then will be having a size of the seal text image in the minimum outsourcing rectangle of (w, h)
It is cut into (h-h%m) w/m2A image block；
If w is not divided exactly by m, and h is not divided exactly by m, then will be having a size of the seal text diagram in the minimum outsourcing rectangle of (w, h)
As being cut to (h-h%m) (w-w%m)/m2A image block.
5. the method according to claim 1, wherein being removed to the seal in each image block to obtain pair
The plain text image for the seal bottom answered, including
By the second target detection model based on Super-resolution reconstruction established model IDN, the seal in each image block is gone
Divided by obtaining the plain text image of corresponding seal bottom.
6. according to the method described in claim 5, it is characterized in that, passing through the second mesh based on Super-resolution reconstruction established model IDN
Detection model is marked, before being removed to the seal in each image block to obtain the plain text image of corresponding seal bottom,
Further include:
Pass through loss function LMAEDriving training, obtains the second target detection mould accordingly based on Super-resolution reconstruction established model IDN
Type:
Wherein, N=m*m*3, IiFor corresponding pixel value at the point i in current image to be reconstructed, Ii ’For the point i in label image
Locate corresponding pixel value, current image to be reconstructed is the image that the second target detection model prediction goes out, and the label image is institute
State the effect image that current image to be reconstructed should be actually reconstructed into.
7. the method according to claim 1, wherein correspondence image block is original in the minimum outsourcing rectangle
Position, by each plain text image mosaic be remove seal after text image in,
If two adjacent image blocks have overlapping in the minimum outsourcing rectangle, overlapping described in the text image after removing seal
Partial pixel value, with the pixel of the corresponding plain text image of adjacent image block to sort rearward in the minimum outsourcing rectangle
Subject to value.
8. a kind of seal eliminating equipment, which is characterized in that the equipment includes:
Seal region acquisition device obtains in the original text image for obtaining the original text image of seal to be removed
Seal region minimum outsourcing rectangle；
Seal cuts removal splicing apparatus, for the seal text image in the minimum outsourcing rectangle to be cut to multiple images
Block；Seal in each image block is removed to obtain the plain text image of corresponding seal bottom；By correspondence image block
Home position in the minimum outsourcing rectangle, is the text image after removing seal by each plain text image mosaic；
Text image after the removal seal spliced is replaced the minimum outsourcing square in the original text image by alternative
Seal region in shape.
9. a kind of computer-readable medium, is stored thereon with computer-readable instruction, which is characterized in that the computer-readable finger
Order can be executed by processor to realize method described in any one of claims 1 to 7.
10. a kind of equipment for handling in network equipment client information, the equipment include for storing computer program instructions
Memory and processor for executing program instructions, which is characterized in that when the computer program instructions are executed by the processor
When, trigger method described in any one of equipment perform claim requirement 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910802744.8A CN110533748A (en) | 2019-08-28 | 2019-08-28 | Seal minimizing technology and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910802744.8A CN110533748A (en) | 2019-08-28 | 2019-08-28 | Seal minimizing technology and equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110533748A true CN110533748A (en) | 2019-12-03 |
Family
ID=68664864
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910802744.8A Pending CN110533748A (en) | 2019-08-28 | 2019-08-28 | Seal minimizing technology and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110533748A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111275051A (en) * | 2020-02-28 | 2020-06-12 | 上海眼控科技股份有限公司 | Character recognition method, character recognition device, computer equipment and computer-readable storage medium |
| CN112069973A (en) * | 2020-09-01 | 2020-12-11 | Pfu上海计算机有限公司 | Seal processing method, device, equipment and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015011684A1 (en) * | 2013-07-26 | 2015-01-29 | G.D. S.P.A. | Inspection method of a product in a packaging machine |
| CN109886974A (en) * | 2019-01-28 | 2019-06-14 | 北京易道博识科技有限公司 | A kind of seal minimizing technology |
-
2019
- 2019-08-28 CN CN201910802744.8A patent/CN110533748A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015011684A1 (en) * | 2013-07-26 | 2015-01-29 | G.D. S.P.A. | Inspection method of a product in a packaging machine |
| CN109886974A (en) * | 2019-01-28 | 2019-06-14 | 北京易道博识科技有限公司 | A kind of seal minimizing technology |
Non-Patent Citations (1)
| Title |
|---|
| 葛怀东等: "古籍书影中钤印提取技术的探讨", 《计算机应用与软件》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111275051A (en) * | 2020-02-28 | 2020-06-12 | 上海眼控科技股份有限公司 | Character recognition method, character recognition device, computer equipment and computer-readable storage medium |
| CN112069973A (en) * | 2020-09-01 | 2020-12-11 | Pfu上海计算机有限公司 | Seal processing method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US8989437B2 (en) | Salient object detection by composition | |
| US8917935B2 (en) | Detecting text using stroke width based text detection | |
| CN111681273B (en) | Image segmentation method and device, electronic equipment and readable storage medium | |
| RU2697649C1 (en) | Methods and systems of document segmentation | |
| US8442327B2 (en) | Application of classifiers to sub-sampled integral images for detecting faces in images | |
| CN110827292B (en) | Video instance segmentation method and device based on convolutional neural network | |
| CN110210400A (en) | A kind of form document detection method and equipment | |
| US20150213330A1 (en) | Methods and systems for efficient automated symbol recognition | |
| EP1730694B1 (en) | Detecting sky occlusion objects in color digital images | |
| EP2680226A1 (en) | Temporally consistent superpixels | |
| CN110533748A (en) | Seal minimizing technology and equipment | |
| CN110059539A (en) | A kind of natural scene text position detection method based on image segmentation | |
| CN105828081A (en) | Encoding method and encoding device | |
| CN110287831A (en) | A kind of acquisition methods, device and the electronic equipment at the control point based on terrestrial reference | |
| CN114677596A (en) | Remote sensing image ship detection method and device based on attention model | |
| CN111310759A (en) | Target detection suppression optimization method and device for dual-mode cooperation | |
| CN114359932B (en) | Text detection method, text recognition method and device | |
| EP2058741A2 (en) | A method and apparatus for analysing a plurality of stored images | |
| US8306111B2 (en) | Image analysis using a hybrid connected component labeling process | |
| EP2966613A1 (en) | Method and apparatus for generating a super-resolved image from an input image | |
| CN116486077B (en) | Remote sensing image semantic segmentation model sample set generation method and device | |
| US8121437B2 (en) | Method and apparatus of searching for images | |
| CN111062385A (en) | Network model construction method and system for image text information detection | |
| CN110019978A (en) | A kind of data processing method and device | |
| SE506145C2 (en) | Method for converting a grayscale image into a black and white image divided into pixels |
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 | ||
| AD01 | Patent right deemed abandoned |
Effective date of abandoning: 20230728 |
|
| AD01 | Patent right deemed abandoned |