CN107506746A

CN107506746A - Locating point-free image identification method and system for intelligent marking system

Info

Publication number: CN107506746A
Application number: CN201710807657.2A
Authority: CN
Inventors: 黎冬媛; 朱春媚; 邹昆; 周文辉
Original assignee: University of Electronic Science and Technology of China Zhongshan Institute
Current assignee: University of Electronic Science and Technology of China Zhongshan Institute
Priority date: 2017-09-08
Filing date: 2017-09-08
Publication date: 2017-12-22

Abstract

The invention provides a method and a system for identifying an image without a positioning point of an intelligent marking system, which relate to the technical field of image processing, and the method comprises the following steps: obtaining a test paper sample to be processed; the method comprises the following steps of learning and processing a test paper sample to obtain a learning result, wherein the learning result comprises: examination number region information, objective question region information and subjective question region information of the test paper sample; acquiring a target test paper to be processed; based on the learning result, adaptively adopting at least one algorithm in a plurality of processing algorithms to identify the target test paper to obtain an identification result of the target test paper, wherein the identification result comprises the following information: the examination number region information of the target test paper, the objective question region information of the target test paper and the subjective question region information of the target test paper are used for solving the technical problem that the accurate positioning and identification of the test paper image cannot be realized by adopting various positioning methods in a self-adaptive manner in the prior art.

Description

Intelligently reading system is without anchor point image-recognizing method and system

Technical field

The present invention relates to the technical field of image procossing, knows more particularly, to a kind of intelligently reading system without positioning dot image Other method and system.

Background technology

With the development of internet, digitlization has become a part indispensable in people's life.Pursue it is quick, Efficiently, under the premise of accurate overall background, the group signature of traditional approach can not gradually meet the needs of people are growing. The mode of group signature, efficiency is low, and easily judges by accident, and papery paper is also required to consume substantial amounts of people in transmitting procedure Power material resources.Therefore, digitlization marking system becomes a kind of inevitable trend.Paper is preserved into digitized map by scanner Picture, then it is handled with computer program, efficiency of going over examination papers can be greatly promoted.And in traditional paper, it is divided into according to type Two major classes：First, there is anchor point paper, this paper is determined by the small square or triangle pair topic on four corners Position, so as to realize objective item identification and subjective item cutting；Second, the paper without positioning, the paper of this type does not have determining for fixation Site, it is different because of paper, have no idea to position it with traditional positioning method, it is therefore desirable to find other positioning side Method.

In view of the above-mentioned problems, do not propose effective solution also.

The content of the invention

In view of this, it is an object of the invention to provide a kind of intelligently reading system without anchor point image-recognizing method and to be System, adaptively it can not be realized with alleviating present in prior art using a variety of localization methods to the accurate fixed of paper image Position and the technical problem of identification.

In a first aspect, the embodiments of the invention provide a kind of intelligently reading system without anchor point image-recognizing method, including： Obtain pending paper sample；Study processing is carried out to the paper sample, obtains learning outcome, wherein, the study knot Fruit includes：The number of the examining area information of the paper sample, objective item area information, subjective item area information；Obtain pending Target paper；Based on the learning outcome, adaptively using at least one of a variety of Processing Algorithms algorithm to the target Paper is identified, and obtains the recognition result of the target paper, wherein, the recognition result includes following information：The mesh Mark the number of the examining area information of paper, the objective item area information of the target paper, the subjective item region letter of the target paper Breath.

Further, obtaining pending paper sample includes：Binary conversion treatment is carried out to original paper sample, obtains two The original paper sample after value processing；Correction processing is carried out to the paper sample after binary conversion treatment, obtained The original paper sample after to correction, and using the original paper sample after correction as the pending examination Roll up sample.

Further, study processing is carried out to the paper sample, obtaining learning outcome includes：In the paper sample At least one localization region is found, wherein, the localization region includes anchor point；Based at least one localization region Anchor point learns to the paper sample, obtains the learning outcome.

Further, the learning outcome includes the number of the examining area information, based at least one localization region Anchor point learns to the paper sample, and obtaining the learning outcome includes：In at least one localization region not In the case of the number of examining region, the first positioning instruction that user sends is obtained, wherein, first positioning instruction is described The instruction of the first template image is positioned in paper sample；After first positioning instruction is got, the use is monitored in real time The region of family institute frame choosing, and using the region of user institute frame choosing as first template image；Getting the user After the first frame choosing instruction sent, into the first mode in the number of examining region described in frame choosing；In the first mode, supervise in real time Survey the number of the examining region of the subscriber frame choosing, and the information in the number of examining region described in record.

Further, the learning outcome includes the objective item area information, based at least one localization region Anchor point the paper sample is learnt, obtaining the learning outcome includes：In at least one localization region Not include objective item region in the case of, obtain user send the second positioning instruction, wherein, second positioning instruction be The instruction of the second template image is positioned in the paper sample；After second positioning instruction is got, institute is monitored in real time The region of user institute frame choosing is stated, and using the region of user institute frame choosing as second template image；It is described getting After the second frame choosing instruction that user sends, the second mode in the objective item region is selected into frame；In the second mode, The objective item region of the subscriber frame choosing is monitored in real time, and records the information in the objective item region.

Further, the learning outcome includes the subjective item area information, based at least one localization region Anchor point the paper sample is learnt, obtaining the learning outcome includes：In at least one localization region In the case of including subjective item region, the second frame choosing instruction of user's transmission is got；In get that the user sends the After the choosing instruction of two frames, the 3rd pattern in the subjective item region is selected into frame；Under the 3rd pattern, in real time described in monitoring The subjective item region of subscriber frame choosing, and record the information in the subjective item region.

Further, finding at least one localization region in the paper sample includes：At following at least one Adjustment method finds at least one localization region in the paper sample：Template matching algorithm and rectangle recognizer.

Further, correction processing is carried out to the paper sample after binary conversion treatment, the institute after being rectified a deviation Stating original paper sample includes：Contours extract is carried out to the original paper model, obtains at least one profile；It is described at least The maximum objective contour of area is selected in one profile；Boundary rectangle processing is carried out to the objective contour, processing obtains described The angle of inclination of objective contour；Rotation process is carried out to the original paper sample based on the angle of inclination, obtains rectifying a deviation it The original paper sample afterwards.

Further, based on the learning outcome, adaptively using at least one of a variety of Processing Algorithms algorithm pair The target paper is identified, and obtains the recognition result of the target paper and includes：Obtain the target in the learning outcome Title field；The Search and Orientation point in the target title field；In the case where not finding the anchor point, lookup and institute State the adjacent anchor point in target title field；By the Processing Algorithm, and determined based on the adjacent anchor point to be identified The rectangular information of title field, and the rectangular information of the title field to be identified is cut.

With reference to second aspect, the embodiments of the invention provide a kind of intelligently reading system without anchor point image identification system, Including：First acquisition unit, for obtaining pending paper sample；Unit, for the paper sample Habit processing, obtains learning outcome, wherein, the learning outcome includes：The number of the examining area information of the paper sample, objective item area Domain information, subjective item area information；Second acquisition unit, for obtaining pending target paper；Recognition unit, for based on The learning outcome, adaptively the target paper is identified using at least one of a variety of Processing Algorithms algorithm, The recognition result of the target paper is obtained, wherein, the recognition result includes following information：The number of the examining area of the target paper Domain information, the objective item area information of the target paper, the subjective item area information of the target paper.

In embodiments of the present invention, paper sample pending first, then, study processing is carried out to paper sample, obtained To learning outcome, wherein, learning outcome includes the number of the examining area information of paper model, objective item area information, subjective item region Information；Next, can obtains pending target paper, and target paper is identified based on learning outcome, obtained The recognition result of target paper, wherein, recognition result includes the number of the examining area information of target paper, the objective item of target paper Area information, the subjective item area information of target paper.In embodiments of the present invention, for without anchor point paper, by first Paper model is learnt target paper to be identified, thus, it is possible to accurately to place is identified without anchor point paper Reason, and then alleviate adaptively can not use a variety of localization methods to realize to the accurate of paper image present in prior art Positioning and the technical problem of identification.

Other features and advantages of the present invention will illustrate in the following description, also, partly become from specification Obtain it is clear that or being understood by implementing the present invention.The purpose of the present invention and other advantages are in specification, claims And specifically noted structure is realized and obtained in accompanying drawing.

To enable the above objects, features and advantages of the present invention to become apparent, preferred embodiment cited below particularly, and coordinate Appended accompanying drawing, is described in detail below.

Brief description of the drawings

, below will be to specific in order to illustrate more clearly of the specific embodiment of the invention or technical scheme of the prior art The required accompanying drawing used is briefly described in embodiment or description of the prior art, it should be apparent that, in describing below Accompanying drawing is some embodiments of the present invention, for those of ordinary skill in the art, before creative work is not paid Put, other accompanying drawings can also be obtained according to these accompanying drawings.

Fig. 1 is a kind of flow chart of the intelligently reading system without anchor point image-recognizing method according to embodiments of the present invention；

Fig. 2 is a kind of schematic diagram of the learning outcome of localization region according to embodiments of the present invention；

Fig. 3 is a kind of schematic diagram of the display interface of frame lectotype according to embodiments of the present invention；

Fig. 4 is a kind of schematic diagram of the learning outcome in the number of examining region according to embodiments of the present invention；

Fig. 5 is a kind of schematic diagram of the learning outcome in objective item region according to embodiments of the present invention；

Fig. 6 is according to embodiments of the present invention a kind of for the line number of setting options and the schematic diagram of columns；

Fig. 7 is the schematic diagram of option learning outcome in a kind of objective item region according to embodiments of the present invention；

Fig. 8 is a kind of schematic diagram of the learning outcome in subjective item region according to embodiments of the present invention；

Fig. 9 is a kind of schematic diagram of the intelligently reading system without anchor point image identification system according to embodiments of the present invention.

Embodiment

To make the purpose, technical scheme and advantage of the embodiment of the present invention clearer, below in conjunction with accompanying drawing to the present invention Technical scheme be clearly and completely described, it is clear that described embodiment is part of the embodiment of the present invention, rather than Whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art are not making creative work premise Lower obtained every other embodiment, belongs to the scope of protection of the invention.

Embodiment one：

According to embodiments of the present invention, there is provided a kind of embodiment of the intelligently reading system without anchor point image-recognizing method, It should be noted that can be in the department of computer science of such as one group computer executable instructions the flow of accompanying drawing illustrates the step of Performed in system, although also, show logical order in flow charts, in some cases, can be with different from herein Order perform shown or described step.

Fig. 1 is a kind of flow chart of the intelligently reading system without anchor point image-recognizing method according to embodiments of the present invention, As shown in figure 1, this method comprises the following steps：

Step S102, obtain pending paper sample；

Step S104, study processing is carried out to the paper sample, obtains learning outcome, wherein, the learning outcome bag Include：The number of the examining area information of the paper sample, objective item area information, subjective item area information；

Step S106, obtain pending target paper；

Step S108, based on the learning outcome, adaptively using at least one of a variety of Processing Algorithms algorithm pair The target paper is identified, and obtains the recognition result of the target paper, wherein, the recognition result includes following letter Breath：The number of the examining area information of the target paper, the objective item area information of the target paper, the subjectivity of the target paper Inscribe area information.

It should be noted that in embodiments of the present invention, it can be performed by the paper ocr software researched and developed in advance.Tool Body, user can be learnt by the software to paper sample, and paper sample is entered for example, manually manipulating the software Row study, or, pass through automatically controlling to realize the study to paper model for the software.Learned in advance to paper sample After habit, it is possible to processing is identified to target paper by the software, is identified result.

In an optional embodiment, obtain pending paper sample and comprise the following steps：

Step S1021, binary conversion treatment is carried out to original paper sample, obtains the original examination after binary conversion treatment Roll up sample；

Step S1022, correction processing is carried out to the paper sample after binary conversion treatment, the institute after being rectified a deviation Original paper sample is stated, and using the original paper sample after correction as the pending paper sample.

In embodiments of the present invention, first, the binary-state threshold of paper sample is adjusted, i.e., two are carried out to original paper sample Value is handled.The general digital paper preserved on computers is all colored or grey, is unfavorable for handling paper, two-value Paper after change can highlight feature, reduce interference.The method of binaryzation is the binaryzation threshold that user rule of thumb adjusts paper Value.

After being scanned to original paper sample, original paper sample occurs different degrees of inclination, therefore, it is necessary to Correction processing is carried out to original paper sample, to improve the accuracy of identification.In embodiments of the present invention, can be to paper sample After this progress binary conversion treatment, it is possible to carry out correction processing to the paper sample after binary conversion treatment, obtain at correction Original paper sample after reason, and using the original paper sample after correction as pending paper sample.

It should be noted that in embodiments of the present invention, to including multiple papers due to a paper, therefore, right When paper sample performs the step described by above-mentioned steps S1021 and step S1022.Need to use above-mentioned steps S1021 successively Study processing is carried out to every paper with the method described by step S1022.

The method of correction processing has many kinds, in embodiments of the present invention, can use largest contours correction or straight line Fitting correction mode carries out correction processing to original paper sample.Below, this two kinds correction processing modes will specifically be introduced.

Mode one, largest contours correction

In an optional embodiment, correction processing is carried out to the paper sample after binary conversion treatment, obtained The original paper sample after to correction comprises the following steps：

First, contours extract is carried out to the original paper model, obtains at least one profile；

Then, the maximum objective contour of area is selected at least one profile；

Next, carrying out boundary rectangle processing to the objective contour, processing obtains the angle of inclination of the objective contour；

Finally, rotation process, the institute after being rectified a deviation are carried out to the original paper sample based on the angle of inclination State original paper sample.

Specifically, in embodiments of the present invention, first, gray processing and two-value are carried out to the scan image of original paper sample The pretreatment of change, and contours extract is carried out to original paper sample, find all outermost profiles in original paper sample (that is, above-mentioned at least one profile).Compare the size of all outermost layer profiles (that is, above-mentioned at least one profile), choose The objective contour of maximum area, the objective contour is carried out to find boundary rectangle processing, the target wheel of maximum may finally be obtained Wide angle of inclination, the angle is exactly the angle of inclination of paper sample.Using the angle of inclination, paper sample can be revolved Turn operation, finally give the paper sample image after correction.

In embodiments of the present invention, the software can realize the lookup of maximum rectangular profile using Opencv functions.Its In, maximum rectangular profile function is found in Opencv is：

intcvFindContours(CvArr*image,CvMemStorage*storage,CvSeq**first_ Contour, intheader_size=sizeof (CvContour), intmode=CV_RETR_LIST, intmethod= CV_CHAIN_APPROX_SIMPLE, CvPointoffset=cvPoint (0,0)).

Image is expressed as：The source bianry image of 8 bit singles, gray level image.Non-zero pixels are as 1 processing, 0 pixel Continue to have.The function for obtaining bianry image from a gray level image has：CvThreshold, cvAdaptiveThreshold and cvCanny。

Storage is expressed as：Return to the container of profile.

First_contour is expressed as：Output parameter, first external profile is pointed to for storing.

Header_size is expressed as：The size of header sequences.If selecting method=CV_CHAIN_CODE, header_size>=sizeof (CvChain)；Other, then header_size>=sizeof (CvContour).

Mode is expressed as：Search modes, can value it is as follows：

CV_RETR_EXTERNAL：Only retrieve outmost profile；

CV_RETR_LIST：All profiles are retrieved, and are put it into list；

CV_RETR_CCOMP：All profiles are retrieved, and they are organized as two layers：Top layer is the external edge of each several part Boundary, the second layer are the borders in cavity；

CV_RETR_TREE：All profiles are retrieved, and reconstruct the whole level of nested profile.

Here we only need exterior contour, therefore select CV_RETR_EXTERNAL.

Method is expressed as：(except the approximation built in CV_RETR_RUNS uses, other patterns make edge approximation method With the approximate data of this setting).

Offset is expressed as：Offset, for moving all profile points.

Mode two, fitting a straight line correction mode

Fitting a straight line correction mode is applied to outside paper answer area unclear in the absence of obvious rectangular shaped rim, frame fracture Clear paper sample, specifically, the specific implementation principles illustrated of fitting a straight line correction mode are as follows：

By way of down-sampling, paper sample as several points, is scanned in longitudinal direction from left to right every fixed, when running into first Stop scanning during individual black picture number point, record the coordinate of the black picture number point.The operation is repeated, until whole paper scanned one It is secondary, one group of coordinate will be now obtained, this group of coordinate is ranked up with the size of abscissa, rejects the coordinate at both ends, to remaining Coordinate carry out least square line fitting, can thus obtain straight line, the angle of inclination of the straight line i.e. paper Angle of inclination, using the angle, rotation process is carried out to paper, with regard to the paper sample after being rectified a deviation.

In another optional embodiment, study processing is carried out to the paper sample, obtaining learning outcome includes：

At least one localization region is found in the paper sample, wherein, the localization region includes anchor point；

Anchor point based at least one localization region learns to the paper sample, obtains the study knot Fruit.

In embodiments of the present invention, after binary conversion treatment and correction processing are carried out to original paper sample, it is possible to The paper model to be analyzed obtained afterwards to processing carries out study processing.

When carrying out study processing to paper sample, for without anchor point paper sample, first without anchor point paper sample At least one localization region is found in this, wherein, the localization region can be understood as the rectangle that can be positioned.Searching out After localization region, the location information in record location region.As shown in Fig. 2 the rectangle frame shown in symbol 1, symbol 2 and symbol 3 The region of institute's frame choosing is localization region.

It should be noted that in embodiments of the present invention, include anchor point in the localization region, i.e. the rectangle positioning area The center in domain is the anchor point of the localization region.After anchor point is determined, it is possible to based on anchor point to paper sample Learnt, obtain the number of examining region, objective item region and subjective item region.

In another optional embodiment, in the case where learning outcome includes the number of the examining area information, it is based on The anchor point of at least one localization region learns to the paper sample, and obtaining the learning outcome includes following step Suddenly：

Step S11, in the case of not including the number of examining region at least one localization region, obtain what user sent First positioning instruction, wherein, first positioning instruction is the instruction that the first template image is positioned in the paper sample；

Step S12, after first positioning instruction is got, the region of user institute frame choosing is monitored in real time, and Using the region of user institute frame choosing as first template image；

Step S13, after the first frame choosing instruction that the user sends is got, into the number of examining region described in frame choosing First mode；

Step S14, in the first mode, the number of the examining region of the subscriber frame choosing is monitored in real time, and examined described in record The information in number region.

In embodiments of the present invention, when learning to paper sample, user can select different in the software Mode of learning, for example, as shown in figure 3, " number of examining " can be selected in interface as shown in Figure 3, " multiple-choice question " and " subjective item ", When user selects " number of examining ", the mode of learning in the number of examining region will be entered.

Before the number of the examining region in sample sample learns, and do not include the number of examining at least one localization region In the case of region, user selects the first template image in the software.Specifically, user can be at interface as shown in Figure 3 Middle selection " template ", now, as send the first positioning instruction to the software.Now, the software will monitor user institute frame in real time The region of choosing.

For example, the frame favored area of " student status full-filling area " these words as shown in Figure 4 be the first template image, it is necessary to Illustrate, before non-frame selects " student status full-filling area " these words, the frame favored area of " student status full-filling area " these words It is not present.User can click on mouse in the upper left corner of " student status full-filling area " these words, and " student status number is filled out Mouse is clicked in the lower right corner of these words of painting area ", now, the position coordinates that the software is clicked on automatic record user, Then, " student status full-filling area " these words are carried out by frame choosing based on the position coordinates, obtains the first template image.

After being positioned to the first template image, it is possible to which the frame for carrying out the number of examining region selects the stage.Now, Yong Huke To click on " number of examining " in interface as shown in Figure 3, now it can be understood that to send the choosing instruction of the first frame to the software.This is soft Part is after the choosing instruction of the first frame is got, it is possible to real into the first mode in the frame choosing number of examining region, and in the flrst mode When monitor the number of the examining region of user institute frame choosing, and record the information in the number of examining region.

Before the frame choosing number of examining region, first have to pre-set first mode, wherein, the setting up procedure of the first mode is retouched State as follows：

Firstly the need of a best orientation distance is determined, the orientation distance of the best orientation distance is arranged to paper length 1/3.When subscriber frame choosing be the number of examining when, judge whether include two anchor points in the best orientation scope in the number of the examining region, if There is no anchor point or only an anchor point in the range of being somebody's turn to do, then prompt user to need to add new anchor point；If in the range of being somebody's turn to do There are two anchor points, can now ensure the precision of positioning.

In interface as shown in Figure 3, when user clicks on " number of examining ", then into first mode.Now, user's can frame The number of examining region is selected, as shown in figure 4, the number of examining region includes 10 column of figures, preceding 4 column of figure of 10 column of figure is to be selected by frame The number of examining region, in 10 column of figures after 6 column of figures be the number of the examining region do not selected by frame also.The number of examining do not selected with the 5th row by frame also Illustrated exemplified by region.User can click on the upper left corner of the column of figure by mouse, and click on the column of figure by mouse The lower right corner.Now, that is, the number of the examining region of user institute frame choosing is monitored, now, based on first mode, the software is by automatic decision Whether two anchor points are included in the range of the best orientation distance in the number of examining region, if including record active user institute frame The number of the examining region of choosing, if do not included, user is prompted to add new anchor point.

It should be noted that in embodiments of the present invention, for every column of figure in as shown in Figure 4, according to above-mentioned institute The method of description is selected and recorded to carry out the frame in the number of examining region.

In another optional embodiment, in the case where learning outcome includes the objective item area information, base Anchor point at least one localization region learns to the paper sample, obtains the learning outcome including as follows Step：

Step S21, in the case of not including objective item region at least one localization region, obtain user and send The second positioning instruction, wherein, second positioning instruction is that the instruction of the second template image is positioned in the paper sample；

Step S22, after second positioning instruction is got, the region of user institute frame choosing is monitored in real time, and Using the region of user institute frame choosing as second template image；

Step S23, after the second frame choosing instruction that the user sends is got, the objective item region is selected into frame Second mode；

Step S24, in the second mode, the objective item region of the subscriber frame choosing is monitored in real time, and based on described Objective item area information.

In embodiments of the present invention, when learning in the objective item region to paper sample, user can be in the software Middle selection " multiple-choice question ", when user selects " multiple-choice question ", the mode of learning (that is, second mode) in objective item region will be entered.

Before the objective item region in sample sample learns, and do not include visitor at least one localization region In the case of sight topic region, user selects the second template image in the software.Specifically, user can be as shown in Figure 3 " template " is selected in interface, now, as sends the second positioning instruction to the software.Now, the software will monitor user in real time The region of institute's frame choosing.

For example, the region " following is multiple-choice question answer area " in Fig. 5 shown in arrow is the second template image, it is necessary to illustrate , before the choosing of non-frame " following is multiple-choice question answer area " these words, " following is multiple-choice question answer area " these words Frame favored area is not present.User can click on mouse in the upper left corner of " following is multiple-choice question answer area " these words Mark, and mouse is clicked in the lower right corner of " following is multiple-choice question answer area " these words, now, the software will record automatically The position coordinates that user is clicked on, then, " following is multiple-choice question answer area " these words are carried out by frame based on the position coordinates Choosing, obtains the second template image.

After being positioned to the second template image, it is possible to which the frame for carrying out objective item region selects the stage.Now, user " multiple-choice question " can be clicked in interface as shown in Figure 3, now it can be understood that to send the choosing instruction of the second frame to the software. The software is after the choosing instruction of the second frame is got, it is possible to the second mode in objective item region is selected into frame, and in the second mould Under formula, the objective item region of user institute frame choosing is monitored in real time, and records the information in objective item region.

Before frame selects objective item region, first have to pre-set second mode, wherein, the setting up procedure of the second mode It is described as follows：

Firstly, it is necessary to determine a best orientation distance, the orientation distance of the best orientation distance is arranged to paper length 1/3.When subscriber frame choosing be multiple-choice question when, whether judge in the best orientation scope in the objective item region comprising two positioning Point, if not having anchor point or only an anchor point in the range of being somebody's turn to do, user is prompted to need to add new anchor point；If the model There are two anchor points in enclosing, can now ensure the precision of positioning.

In interface as shown in Figure 3, when user clicks on " multiple-choice question ", then into second mode.Now, user's can Frame selects objective item region, as shown in figure 5, the region shown in symbol 4 is the approximate region in objective item region.One has been selected in frame After approximate region, the line number and columns of option are inserted in interface as shown in Figure 6, the software must detect this substantially by automatic The position of each option and size in region, the schematic diagram for detecting the obtained position of each option and size are as shown in Figure 7.

It is as follows to implement principles illustrated：By as the method gone through several times detect in approximate region the position of option with it is big It is small, carried out from the surrounding of option toward center as the scanning of several points, if a number of black color dots picture number is run into, stop sweeping Retouch.Four end points of option can be thus obtained, so that it is determined that the position of option and size.

In another optional embodiment, include the situation of the subjective item area information in the learning outcome Under, the anchor point based at least one localization region learns to the paper sample, obtains the learning outcome bag Include following steps：

Step S31, in the case of including subjective item region at least one localization region, get user's transmission The second frame choosing instruction；

Step S32, after the second frame choosing instruction that the user sends is got, the subjective item region is selected into frame The 3rd pattern；

Step S33, under the 3rd pattern, the subjective item region of the subscriber frame choosing is monitored in real time, and based on described Subjective item area information.

In embodiments of the present invention, when learning in the objective item region to paper sample, user can be as shown in Figure 3 Interface in select " subjective item ", when user selects " subjective item ", the mode of learning the (that is, the 3rd in subjective item region will be entered Pattern).

Before the objective item region in sample sample learns, and include at least one localization region it is objective In the case of inscribing region, user can click on " subjective item " in interface as shown in Figure 3, now it can be understood that being soft to this Part sends the choosing instruction of the 3rd frame.The software is after the choosing instruction of the 3rd frame is got, it is possible to selects subjective item region into frame 3rd pattern, and in a third mode, the subjective item region of user institute frame choosing is monitored in real time, and record the letter in subjective item region Breath.

Before frame selects objective item region, first have to pre-set the 3rd pattern, wherein, the setting up procedure of the 3rd pattern It is described as follows：

Firstly the need of a best orientation distance is determined, the orientation distance of the best orientation distance is arranged to paper length 1/3.When subscriber frame choosing be subjective item when, whether judge in the best orientation scope in the subjective item region comprising a positioning Point, if not having anchor point in the range of being somebody's turn to do, user is prompted to need to add new anchor point；If having an anchor point in the range of being somebody's turn to do, The precision of positioning can now be ensured.

In interface as shown in Figure 3, when user clicks on " subjective item ", then into the 3rd pattern.Now, user's can Frame selects subjective item region.Specifically, a subjective item region of as frame choosing as shown in Figure 8.Master as shown in Figure 8 is selected in frame Before sight topic region, user can click on the upper left corner in the region by mouse, and the lower right corner in the region is clicked on by mouse, After position coordinates of the software detection to two points, it becomes possible to determine a subjective item region based on the position coordinates.

In another optional embodiment, in the paper sample finding at least one localization region includes：It is logical Cross following at least one method and at least one localization region is found in the paper sample：Template matches are calculated and rectangle is known Other algorithm.

In another optional embodiment, based on the learning outcome pair, adaptively using a variety of Processing Algorithms At least one of target paper described in algorithm be identified, the recognition result for obtaining the target paper comprises the following steps：

Step S1081, obtain the target title field in the learning outcome；

Step S1082, the Search and Orientation point in the target title field；

Step S1083, in the case where not finding the anchor point, search and the target title field is adjacent determines Site；

Step S1084, title field to be identified is determined by the Processing Algorithm, and based on the adjacent anchor point Rectangular information, and the rectangular information of the title field to be identified is cut.

Specifically, in embodiments of the present invention, when target paper is identified, paper is rectified a deviation first, entangled Inclined method includes two kinds, first, the method for correcting error of fitting a straight line, user selects the straight line portion on paper, and then, sampling is intended Straight line is closed out, calculates the angle of inclination of straight line, then paper is rectified a deviation；Second, the side by finding maximum rectangle Method, a rectangle maximum on paper is found, the angle of inclination of the rectangle is calculated, is then rectified a deviation.Next, it is judged that target Whether the page number of paper is accurate, specifically, can contrast by the page number identification region recorded in learning database and judge the paper page number It is whether correct.After contrast is correct, it is possible to take out target title field, whether is the anchor point checked in target title field It has been found that, if not finding, calculate and find out the anchor point progress fixation and recognition nearest apart from the region, find positioning After point, the mark of the anchor point is designated as " identification ", other zone locations use after being provided with, without again fixed again Position.After finding a positioning, the rectangular information of target title field on paper is calculated, will be main if topic is subjective item Sight topic is cut out；If topic is objective item (including the number of examining), the option of full-filling is identified.If whole anchor point identifications are lost Lose, then labeled as the paper of mistake；Each paper, and output result are handled successively according to above step.

It should be noted that by foregoing description, in embodiments of the present invention, can use template matching algorithm and Paper sample localization region is identified rectangle recognizer.Above two algorithm can be also used for determine in target paper Rectangular area.

For example, template matching algorithm, is that the sub-fraction (for example, target title field) chosen on paper sample is used as mould Plate, the same place on target paper to be identified is found by template matching algorithm, so as to which the rectangular information for finding similar enters Row positioning.Rectangle recognizer, learn to paper sample, after obtaining learning outcome, it is possible to based on learning outcome The positioning and identification of rectangular information are carried out to target paper to be identified.

Below, above two algorithm will be described in detail.

Method one, the calculating process of template matching algorithm are described as follows：

During learning to paper sample, the process that is learnt using template matching algorithm to paper sample It is described as follows：

First, after user selects template image on paper sample, the dutycycle of calculation template image, if dutycycle is very few Then give up the template, prompt user to choose again, until template dutycycle reaches satisfactory degree.

Then, the position of logging template image, wide height, and using wide high expansion 20% as matching region of search, and run CvMatchTemplate, optimal Similarity value SMax1 of the logging template image in region of search is matched.

During target paper is identified, target paper is identified using template matching algorithm process It is described as follows：

1) best match degree threshold value TMax, at least matching degree threshold value Tmin and matching region of search N are set；

2) template image file, position, wide high, matching region of search are read；

3) cvMatchTemplate is carried out to template image in matching region of search to operate to obtain optimal similarity Value SMax2, similarity S=SMax2/SMax1 is calculated, if S>=TMax, then it is assumed that position successfully, return to template coordinate as fixed Site；If S<Tmax then expands matching region of search 1 time, and it is n that note, which expands number,；

4) " n is judged<N" if so, the similarity maximum SMaxTmp and relevant information in this n times matching are recorded, and Skip to the 3rd step；Otherwise, carry out in next step

5) SMaxTmp is judged>Whether Tmin sets up, if so, the positional information during maximum similarity value is then returned, it is no Then, it fails to match for return.

Method two, rectangle recognizer

The algorithm is divided into two big steps, when on paper sample localization region record, second, on pending target paper The fixation and recognition of rectangular information.Its overall thought is first to remember that the size of localization region, frame line segment are being known on paper sample Ranking index and length in all straight lines not gone out, square is made on pending target paper according to the information of these records The identification of shape information.

Step 1: on paper sample localization region record

First, user's substantially frame selects rectangular area, then, identifies the largest contours in the region, and use rectangle fitting； Next, carried out once " the overlapping judgement of rectangle line segment " to rectangle, effect be to ensure that largest contours search out carry out rectangle with it is straight It is consistent that line fits the rectangle size come.

" the overlapping judgement of rectangle line segment " method is as follows：Word in paper can influence the identification of line segment, for mistake as far as possible Filter the pixel of interference, it is necessary first to processing is purged to rectangular area, step is as follows：Determine that the pixel of rectangle frame is wide, General 3~6 pixels, are rule of thumb arranged to 5；From localization region from top to bottom transversal scanning image, 5 are scanned successively downwards Pixel, if wherein there is black pixel point, stop scanning, and total pixel SumPix of the row is added 1；When scan through a line it Afterwards, SumPix and positioning area field width size are compared, if SumPix is more than the 80% of positioning area field width, then it is assumed that the row is possible to It is the length of side of rectangle, and stops scanning；Otherwise, then by the row pixel whole zero setting.Successively from upper and lower, left and right four direction pair Image is purged operation, filters off noise as much as possible.Rectangular region image after removing is handled is become with Hough Exchange the letters number carries out straight-line detection, obtains the line segment information in the region.Because line Segment has the reasons such as thickness, fracture, Hough becomes Change the general effect of the line segment detected be not it is very perfect, such as a line segment will detect that it is a plurality of, it is therefore desirable to detecting The line segment come carries out after-treatment.

Processing step is as follows：1. line segment is divided into four classes, when the line segment on the vertical left side, second, the line segment on vertical the right, Third, the line segment above level, fourth, the line segment of horizontal bottom, sorting technique：The angle of line segment is calculated according to line segment coordinate, then Compare the magnitude relationship of line segment coordinate and regional center, draw the classification of line segment；2. merging line segment, taken from sorted line segment Going out two lines section, calculate their distances on respective direction, vertical line segment distance is the absolute value subtracted each other of y-axis coordinate, Horizontal line segment distance is the absolute value subtracted each other of x-axis coordinate, compares distance whether in 5 pixels, if then straight by two Line merges, and the method for merging is to take the coordinate of the respective end points of two straight lines；3. the line segment after merging is sorted, because to record square Index number of the shape frame line segment in all line segments identified.

Therefore need that the line segment of diverse location is needed to make different sequences：Top line segment：Sorted according to y-axis size；Bottom Portion's line segment：Sorted according to sections bottom positional distance size is found；Left side line segment：Sorted according to x-axis size；Right side line segment：Root Sorted according to region right positions are found apart from size；Record the serial number information of rectangle line segment on four direction.

Step 2: the fixation and recognition of the rectangular information on pending target paper

1. the line segment of area information is detected with Hough transformation, and the line segment to detecting pre-processes, including classification, Merge；

2. a pair pretreated line segment is weighted sequence.Weighted Segments sequence is to be based on such a problem：Hough Conversion will detect that many bar line segments, then just need to select the sideline of rectangle.If all line segments are all tested one time If drawing optimal rectangle, it is assumed that there are 10 line segments in each direction, then fits the fitting required for optimal rectangle Trial is at most likely to be：10^4=10000 times.The efficiency so identified will be very low, it is therefore desirable to which line segment is weighted Sequence, filters out weights highest, that is, the line segment being most likely to be on rectangle sideline, on the contrary it will not be possible to line segment abandon, finally Obtaining three line segments of weights highest, then each fitted rectangle worst is also 3^4=81 times, and in many cases, because To have filtered out weights highest line segment, therefore the general trial for only needing 3-5 time can be to obtain optimal rectangle.Institute It is a very important step with Weighted Segments sequence.Line segment weights include the content of two aspects, first, segment positions error is weighed Value, second, line segment length error weights.Because originally have recorded the sideline sequence number of rectangle in all directions in template, because This is more likely to be the sideline of rectangle closer to the line segment of the sequence number, but because the expansion of template paper and paper to be identified identifies The reason for region is not quite identical, sequence number is identical to be exactly not necessarily the sideline of rectangle, so the judgement of line segment length is also added, The line segment of length rectangle side line length in template is more likely to be the rectangle sideline for needing to find.

Segment positions error weights determine that its calculation formula is by the sequence number difference of the line segment sequence number after sorting and record： Site error weights=0.8^abs (current line segment sequence number-record line segment sequence number).Wherein, when sequence number error be 0, weights 1, When sequence number error increases, weights are reduced rapidly due to power operation, can be significantly the line segment area of each stratum Separate.

Line segment length error weights in order to avoid the excessive influence of error information, it is necessary to error is normalized, Normalized method is min-max standardization.The rectangle side line length error of every line segment and record is calculated first, and preserves mistake The maximin of difference, it is by the normalized formula of error：Error in length weights=(line segment length-minimum line segment length)/(most Big line segment length-minimum line segment length).After normalization, each error information can play identical action effect.It is calculated After the weights of every line segment, line segment is ranked up from big to small according to weights.

3. the line segments extraction that weights are located to front three comes out, remaining is given up.

4. being fitted from a line segment is respectively taken out up and down, the method for fitting is：Obtain line segment and left line segment The intersection point Point2 of intersection point Point1, lower line segment and right line segment, the rectangle length for being fitted to obtain are：Point2.x-Point1.x； Highly it is：Point2.y-Point1.y.

5. compare the rectangle recorded in template to grow tall, if error is within 8 pixels, then it is assumed that identify successfully, return The coordinate of fitted rectangle and grow tall；If the error of length or height wherein has one to exceed 8 pixels, carry out in next step.

6. taking out next candidate line sections in the line segment aggregate of top, it is fitted again.

7. when top line segment aggregate is all completed and does not find suitable rectangle, by the line segment rope of the direction Draw replacement, and the authority for changing line segment is moved into next direction according to the order of " upper, left, down, right ".

It is defeated if suitable rectangle is not found after 8. all line segment aggregates on four direction all test fitting Go out " rectangle is not found ".

It should be noted that in addition to above-mentioned several localization methods, it is also an option that other localization methods come to mesh The rectangular information marked in paper carries out fixation and recognition, for example, it is also possible to using rectangular profile matching algorithm and fitting a straight line rectangle Algorithm carrys out Search and Orientation point.

Because the software is directed to the paper of no anchor point, it is therefore desirable to other localization method is selected paper, should The location algorithm of software selection is template matches+rectangular profile matching algorithm+fitting a straight line rectangle.Three kinds of localization methods can be excessively right Paper is more accurately positioned, and reduces fault rate.

The method that template matches use is correlation matching algorithm, relative value and image pair of this method by template to its average The relative value of its average is matched, and 1 represents perfect matching, and -1 represents the matching of worst, and 0 represents no any correlation. Rectangular profile identification is that larger rectangle is automatically found on paper, and the method by finding outline determines the position of rectangle Size is used as the region of a positioning.Fitting a straight line rectangle is then to fit one by identifying the four edges line segment of rectangle Rectangle, judge whether the rectangle is in the same size with record, if if consistent, a positioning can be turned into by fitting the rectangle come Point.

It is not by paper content constraints that the method for above-mentioned three kinds of positioning, which respectively has the advantages of advantage and disadvantage, template matches, is applied to The paper of most of no anchor points, shortcoming are that speed is slow, and locating effect is had a great influence by paper fog-level；The side of rectangular profile Method advantage is simple and quick, but larger by the restriction of paper content, such as the rectangle of paper has the fracture more than at two to cause to determine Position failure, meanwhile, because being to find outline, if there is stroke protrusion in rectangular edges, also result in rectangle positioning failure；Straight line is intended The method advantage of conjunction is can to resist that paper is fuzzy, rectangle fracture serious situation, but speed is slower, and if rectangle inside Or outside has a plurality of straight line also to be affected to identification.By four kinds of method complementations, location efficiency can be effectively improved And accuracy.

Above-mentioned rectangular profile matching algorithm and fitting a straight line rectangle algorithm is explained below

Method three, rectangular profile matching algorithm

The thought of rectangular profile matching finds in paper sample and target paper wide high identical rectangle as paper Anchor point.The difficult point of this method is in fracture, variable thickness, stroke interference, inclination existing for the rectangle in solution paper is needed Etc. various problems with method be largest contours fitted rectangle.By finding a range of largest contours, then with one Individual minimum rectangle is fitted to it, judges whether the rectangle that this is fitted is wide high consistent with the rectangle in template paper, Using the upper left corner of the rectangle as anchor point if consistent.

(1) profile maximum in subscriber frame favored area is found, and uses rectangle fitting；

(2) according to region of search size 20% couple its be enlarged；

(3) rectangular information is recorded.

(1) identification parameter is set, parameter includes：Region of search expands times N, the wide error acceptance level WE of rectangle, rectangle High error acceptance level HE, profile size filtering M, and initialize and expand region of search frequency n=0；

(2) if n>N, which is then returned, finds rectangle failure；Otherwise, carry out in next step.

(3) all profiles in region of search are found；

(4) profile size is judged>Whether M sets up, if so, then retaining the profile, otherwise, deletes the profile；

(5) profile is fitted with minimum rectangle；

(6) next rectangle will be fitted to sort by size；

(7) maximum rectangle is taken out, the width of the rectangle, height are subtracted each other with the width of rectangle, height in template paper, obtained Wide error W, high error H, if W<=WE and H<=HE, then assert that rectangle is found successfully；Otherwise, region of search is expanded, n adds 1, Return to (2) step.

Method four, fitting a straight line rectangle algorithm

Fitting a straight line rectangle is considered based on such：Although largest contours fitting can recognize that a part of rectangle, Still have the place of deficiency, if such as rectangle have many places fracture that the error of largest contours will be caused very big, or rectangle line segment On there is stroke to be connected to also result in and fit that the error come is very big, and fitting a straight line rectangle can solve these problems well. The basic thought of the algorithm is the line segment that certain area is identified by Hough transformation, the rectangular information recorded on contrast mould's paper So as to fit a similar rectangle of size as anchor point.In embodiments of the present invention, mainly Hough transformation function is used To realize the Algorithm of fitting a straight line.

Hough change function is in Opencv：void HoughLinesP(InputArray image, OutputArray lines,double rho,double theta,int threshold,double minLineLength =0, double maxLineGap=0).

First parameter, the image of InputArray types, input picture, i.e. source images, need to be the single channel two of 8 Enter imaged, after being modified as this form by function after being loaded into by arbitrary source figure, then fill out herein.

Second parameter, the lines of InputArray types, detection is stored after calling HoughLinesP functions The output vector of the lines arrived, each line by with four elements vector (x_1, y_1, x_2, y_2) represent, wherein, (x_ 1, y_1) and (x_2, y_2) is the end point of each line segment detected.

3rd parameter, the rho of double types, the range accuracy in units of pixel.The another kind mode of describing is straight The unit radius of progress size during line search.

4th parameter, the theta of double types, the angle precision in units of radian.The another kind mode of describing is The unit angle of progress size during linear search.

5th parameter, the threshold of int types, the threshold parameter for the plane that adds up, that is, it is in figure to identify certain part The value that it must reach in cumulative plane during straight line.Line segment more than threshold value threshold, which can be just detected, to be passed through And return in result.

6th parameter, the minLineLength of double types, there is default value 0, represents the length of minimum line segment, than The short line segment of this setup parameter cannot be revealed out.

7th parameter, the maxLineGap of double types, there is default value 0, it is allowed to will connect between points with a line Pick up the maximum distance come.

Embodiment two：

The embodiment of the present invention additionally provides a kind of intelligently reading system without anchor point image identification system, the intelligently reading system Unite without anchor point image identification system be mainly used in performing the intelligently reading system that is provided of the above of the embodiment of the present invention without Anchor point image-recognizing method, intelligently reading system provided in an embodiment of the present invention is done without anchor point image identification system below It is specific to introduce.

Fig. 9 is a kind of schematic diagram of the intelligently reading system without anchor point image identification system according to embodiments of the present invention, As shown in figure 9, the intelligently reading system mainly includes without anchor point image identification system：First acquisition unit 91, unit 92, second acquisition unit 93 and recognition unit 94, wherein：

First acquisition unit, for obtaining pending paper sample；

Unit, for carrying out study processing to the paper sample, learning outcome is obtained, wherein, the study knot Fruit includes：The number of the examining area information of the paper sample, objective item area information, subjective item area information；

Second acquisition unit, for obtaining pending target paper；

Recognition unit, for the target paper to be identified based on the learning outcome, obtain the target paper Recognition result, wherein, the recognition result includes following information：The number of the examining area information of the target paper, the target The objective item area information of paper, the subjective item area information of the target paper.

In addition, in the description of the embodiment of the present invention, unless otherwise clearly defined and limited, term " installation ", " phase Even ", " connection " should be interpreted broadly, for example, it may be being fixedly connected or being detachably connected, or be integrally connected；Can To be mechanical connection or electrical connection；Can be joined directly together, can also be indirectly connected by intermediary, Ke Yishi The connection of two element internals.For the ordinary skill in the art, with concrete condition above-mentioned term can be understood at this Concrete meaning in invention.

In the description of the invention, it is necessary to explanation, term " " center ", " on ", " under ", "left", "right", " vertical ", The orientation or position relationship of the instruction such as " level ", " interior ", " outer " be based on orientation shown in the drawings or position relationship, merely to Be easy to the description present invention and simplify description, rather than instruction or imply signified device or element must have specific orientation, With specific azimuth configuration and operation, therefore it is not considered as limiting the invention.In addition, term " first ", " second ", " the 3rd " is only used for describing purpose, and it is not intended that instruction or hint relative importance.

It is apparent to those skilled in the art that for convenience and simplicity of description, the system of foregoing description, The specific work process of device and unit, the corresponding process in preceding method embodiment is may be referred to, will not be repeated here.

In several embodiments provided herein, it should be understood that disclosed systems, devices and methods, can be with Realize by another way.Device embodiment described above is only schematical, for example, the division of the unit, Only a kind of division of logic function, can there is other dividing mode when actually realizing, in another example, multiple units or component can To combine or be desirably integrated into another system, or some features can be ignored, or not perform.It is another, it is shown or beg for The mutual coupling of opinion or direct-coupling or communication connection can be by some communication interfaces, device or unit it is indirect Coupling or communication connection, can be electrical, mechanical or other forms.

The unit illustrated as separating component can be or may not be physically separate, show as unit The part shown can be or may not be physical location, you can with positioned at a place, or can also be distributed to multiple On NE.Some or all of unit therein can be selected to realize the mesh of this embodiment scheme according to the actual needs 's.

In addition, each functional unit in each embodiment of the present invention can be integrated in a processing unit, can also That unit is individually physically present, can also two or more units it is integrated in a unit.

If the function is realized in the form of SFU software functional unit and is used as independent production marketing or in use, can be with It is stored in the executable non-volatile computer read/write memory medium of a processor.Based on such understanding, the present invention The part that is substantially contributed in other words to prior art of technical scheme or the part of the technical scheme can be with software The form of product is embodied, and the computer software product is stored in a storage medium, including some instructions are causing One computer equipment (can be personal computer, server, or network equipment etc.) performs each embodiment institute of the present invention State all or part of step of method.And foregoing storage medium includes：USB flash disk, mobile hard disk, read-only storage (ROM, Read- Only Memory), random access memory (RAM, Random Access Memory), magnetic disc or CD etc. are various can be with The medium of store program codes.

Finally it should be noted that：Embodiment described above, it is only the embodiment of the present invention, to illustrate the present invention Technical scheme, rather than its limitations, protection scope of the present invention is not limited thereto, although with reference to the foregoing embodiments to this hair It is bright to be described in detail, it will be understood by those within the art that：Any one skilled in the art The invention discloses technical scope in, it can still modify to the technical scheme described in previous embodiment or can be light Change is readily conceivable that, or equivalent substitution is carried out to which part technical characteristic；And these modifications, change or replacement, do not make The essence of appropriate technical solution departs from the spirit and scope of technical scheme of the embodiment of the present invention, should all cover the protection in the present invention Within the scope of.Therefore, protection scope of the present invention described should be defined by scope of the claims.

Claims

1. a kind of intelligently reading system is without anchor point image-recognizing method, it is characterised in that including：

Obtain pending paper sample；

Study processing is carried out to the paper sample, obtains learning outcome, wherein, the learning outcome includes：The paper sample This number of examining area information, objective item area information, subjective item area information；

Obtain pending target paper；

Based on the learning outcome, adaptively the target paper is entered using at least one of a variety of Processing Algorithms algorithm Row identification, obtains the recognition result of the target paper, wherein, the recognition result includes following information：The target paper The number of examining area information, the objective item area information of the target paper, the subjective item area information of the target paper.

2. according to the method for claim 1, it is characterised in that obtaining pending paper sample includes：

Binary conversion treatment is carried out to original paper sample, obtains the original paper sample after binary conversion treatment；

Correction processing is carried out to the paper sample after binary conversion treatment, the original paper sample after being rectified a deviation This, and using the original paper sample after correction as the pending paper sample.

3. according to the method for claim 1, it is characterised in that study processing is carried out to the paper sample, learnt As a result include：

Anchor point based at least one localization region learns to the paper sample, obtains the learning outcome.

4. according to the method for claim 3, it is characterised in that the learning outcome includes the number of the examining area information, base Anchor point at least one localization region learns to the paper sample, and obtaining the learning outcome includes：

In the case of not including the number of examining region at least one localization region, obtain the first positioning that user sends and refer to Order, wherein, first positioning instruction is the instruction that the first template image is positioned in the paper sample；

After first positioning instruction is got, the region of user institute frame choosing is monitored in real time, and by the user institute The region of frame choosing is as first template image；

After the first frame choosing instruction that the user sends is got, into the first mode in the number of examining region described in frame choosing；

In the first mode, the number of the examining region of the subscriber frame choosing, and the information in the number of examining region described in record are monitored in real time.

5. according to the method for claim 3, it is characterised in that the learning outcome includes the objective item area information, Anchor point based at least one localization region learns to the paper sample, and obtaining the learning outcome includes：

In the case of not including objective item region at least one localization region, obtain the second positioning that user sends and refer to Order, wherein, second positioning instruction is the instruction that the second template image is positioned in the paper sample；

After second positioning instruction is got, the region of user institute frame choosing is monitored in real time, and by the user institute The region of frame choosing is as second template image；

After the second frame choosing instruction that the user sends is got, the second mode in the objective item region is selected into frame；

In the second mode, the objective item region of the subscriber frame choosing is monitored in real time, and records the objective item region Information.

6. according to the method for claim 3, it is characterised in that the learning outcome includes the subjective item area information, Anchor point based at least one localization region learns to the paper sample, and obtaining the learning outcome includes：

In the case of including subjective item region at least one localization region, the second frame choosing for getting user's transmission refers to Order；

After the second frame choosing instruction that the user sends is got, the 3rd pattern in the subjective item region is selected into frame；

Under the 3rd pattern, the subjective item region of the subscriber frame choosing is monitored in real time, and records the subjective item region Information.

7. according to the method for claim 3, it is characterised in that find at least one localization region in the paper sample Including：

At least one localization region is found in the paper sample by following at least one algorithm：Template matching algorithm With rectangle recognizer.

8. according to the method for claim 2, it is characterised in that the paper sample after binary conversion treatment is entangled Processing, the original paper sample after being rectified a deviation include partially：

Contours extract is carried out to the original paper model, obtains at least one profile；

The maximum objective contour of area is selected at least one profile；

Boundary rectangle processing is carried out to the objective contour, processing obtains the angle of inclination of the objective contour；

Rotation process, the original paper after being rectified a deviation are carried out to the original paper sample based on the angle of inclination Sample.

9. according to the method for claim 1, it is characterised in that based on the learning outcome, adaptively using a variety of places The target paper is identified at least one of adjustment method algorithm, obtains the recognition result of the target paper and includes：

Obtain the target title field in the learning outcome；

The Search and Orientation point in the target title field；

In the case where not finding the anchor point, the anchor point adjacent with the target title field is searched；

By the Processing Algorithm, and the rectangular information of title field to be identified is determined based on the adjacent anchor point, and it is right The rectangular information of the title field to be identified is cut.

10. a kind of intelligently reading system is without anchor point image identification system, it is characterised in that including：

First acquisition unit, for obtaining pending paper sample；

Unit, for carrying out study processing to the paper sample, learning outcome is obtained, wherein, the learning outcome bag Include：The number of the examining area information of the paper sample, objective item area information, subjective item area information；

Second acquisition unit, for obtaining pending target paper；

Recognition unit, for based on the learning outcome, adaptively using at least one of a variety of Processing Algorithms algorithm pair The target paper is identified, and obtains the recognition result of the target paper, wherein, the recognition result includes following letter Breath：The number of the examining area information of the target paper, the objective item area information of the target paper, the subjectivity of the target paper Inscribe area information.