CN106203496A - Hydrographic curve extracting method based on machine learning - Google Patents
Hydrographic curve extracting method based on machine learning Download PDFInfo
- Publication number
- CN106203496A CN106203496A CN201610520993.4A CN201610520993A CN106203496A CN 106203496 A CN106203496 A CN 106203496A CN 201610520993 A CN201610520993 A CN 201610520993A CN 106203496 A CN106203496 A CN 106203496A
- Authority
- CN
- China
- Prior art keywords
- machine learning
- curve
- extracting method
- method based
- hydrographic
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/56—Extraction of image or video features relating to colour
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of hydrographic curve extracting method based on machine learning.When the inventive method is to hydrological data image march line drawing, select and extract some feature in image with resolving ability, and use the sampling window of changeable scale that the image pixel of certain area is sampled, as sample data, divided the iconic element with different characteristic by the method for machine learning, and add new training sample according to classifying quality by incremental mode;And utilize chain code following to carry out post processing, the influence of noise produced after effectively eliminating classification.Compared to existing technology, the present invention solve hydrographic curve to be extracted thinner time especially prominent aim curve disconnection problem, and this problem is difficult to effectively be solved in original hydrographic curve extracting method.
Description
Technical field
The present invention relates to a kind of image extraction method, particularly relate to the extraction side of hydrographic curve in hydrological data image
Method, belongs to image segmentation field.
Background technology
In the most information-based and digitized epoch, along with the universal of computer and the high speed development of storage medium, respectively
Plant research field all the digitized of data message to be paid attention to further.Due to historical reasons, the field such as hydrology and water conservancy is used mostly net
Trrellis diagram paper hourly observation data.But paper material can cause the problems such as damage, pollution owing to preserving the reason such as improper, the most right
The information carried causes damage.And paper material takes up space, it is not easy to again exchange and the transmission of information, more likely buries
That magnanimity information may be hidden, need excavate knowledge.It is therefore desirable to these papery data are digitized.Utilize figure
Mode as processing by these information gatherings and sets up data base, will avoid the substantial amounts of manual duplication of labour, also can be the most accurate
These information are carried out typing, there is stronger actual application value.
Papery hydrological data is typically the hepatic hydrographic curve drawn on the coordinate net ruled paper of Chinese red, in numeral
During change, it is accomplished by each intersection point obtaining hydrographic curve with grid lines when obtaining the information in drawing, as each moment
Observation.This is crossed range request and splits image, relate to grid lines segmentation and splits with hydrographic curve.
Image segmentation becomes several regions specific, with unique properties image according to certain criteria exactly
And therefrom extract technology and the process of interesting target.Image segmentation is the key precondition of graphical analysis, the quality of its segmentation
Quality is largely fixed the effect that successive image is analyzed.Image segmentation can be divided into gray level image segmentation and coloured image to divide
Cut.Compared with gray level image, coloured image not only comprises monochrome information, has further included shades of colour information, and its partitioning scheme is more
For various, but corresponding segmentation difficulty is the biggest.So far, research worker both domestic and external is in color images field
Have been carried out substantial amounts of research, and propose many partitioning algorithms, and the segmentation strategy for specific image, mainly include base
In histogram thresholding method, based on region method, edge detection method, Segmentation by Fuzzy Clustering method and neural network etc..
In research before, the segmentation of hydrological data image is generally used threshold value based on color histogram and divides
Analysis method, it is also considered that gradient information uses with the fusion of colouring information.This type of method can adaptive complete generally
Image segmentation, and to reduce camera shooting be the impact of uneven illumination.But find that when such method actually used extraction obtains
Hydrographic curve the most easily produce broken string at some, and usually disconnected the most serious, it is difficult to use expanding method solution
Certainly.
Summary of the invention
Goal of the invention: for the digitized of papery hydrological data, it is provided that a kind of hydrographic curve extracting method, it is possible to accurately
Extract hydrographic curve therein, effectively evade curve disconnection problem.
The hydrographic curve extracting method of the present invention, involved hydrological data image is by shooting papery hydrological data
Arrive.
The present invention solves the problems referred to above the most by the following technical solutions.
A kind of hydrographic curve extracting method based on machine learning, comprises the following steps:
Step A, the yardstick of selected sampling window and the target characteristic that need to sample, and gather representative training accordingly
Sample set;The yardstick of described window is scalable.The selection of window size determines the data volume for classification, also directly affects
The scale of amount of calculation.
Step B, utilize the method for machine learning train from training sample generation classification forecast model;
Step C, to each pixel in pending image, collect target characteristic as to be sorted according to sampling window
Sample, the classification forecast model utilizing the training of step B to obtain is classified;
Step D, judge that in pending image, the classification results of each pixel is preferable so that curve extracts complete and do not has
There is the obvious region of other classification errors.The most then enter step F;Otherwise, step E is entered;
Step E, from curve broken string region and the obvious region of classification error choose representative sample point, to it
Add to after sampling in training sample set, and repeat step B;
Step F, to process after image carry out post processing, remove noise that may be present.
Preferably, the training sample set gathered in step A should include at least " hydrographic curve ", " grid lines ", " other
Background " sample of three kinds.
Preferably, the post processing of image method used in step F uses chain code following to combine with expansion process.Wherein chain
Before code tracking hydrographic curve, first grid lines is tracked, determine grid lines corresponding make graph region;This step can alleviate with
Process intensity during track hydrographic curve.
Preferably, after chain code following, the connected domain being smaller in size than specific threshold is regarded as noise, and eliminate image.Should
Threshold value value is 10000.
Preferably, the size of connected domain uses the area of minimum enclosed rectangle of this connected domain to represent.
Compared to existing technology, the method have the advantages that
One, the present invention can preferably solve the disconnection problem easily produced when extracting fine rule;
Two, the present invention is based on the classification to sample mode, as long as selecting sufficient training sample, is not required to consider illumination
Impact etc. problem;
Three, use off-line learning, be not required to each image Resurvey sample training.
Accompanying drawing explanation
Fig. 1, Fig. 2 and Fig. 3 are that three width shoot the hydrological data image obtained.
Fig. 4 a and Fig. 4 b is the result that Fig. 1 and Fig. 2 is extracted hydrographic curve by existing method.
Fig. 5 a and Fig. 5 b be in the inventive method the different phase of train classification models Fig. 2 is classified prediction result.
Fig. 6 a and Fig. 6 b be in the inventive method the different phase of train classification models Fig. 3 is classified prediction result.
Fig. 7 a and Fig. 7 b is the result that Fig. 2 and Fig. 3 is extracted hydrographic curve by the inventive method.
Fig. 8 is the flow chart of the present invention.
Detailed description of the invention
Below in conjunction with the accompanying drawings technical scheme is described in detail: Fig. 1 and Fig. 2 respectively illustrates the two width hydrology
Source map picture, it is critical only that hydrographic curve therein (bluish violet) and coordinate net ruling (Chinese red) what it was digitized
Extraction.As can be seen from the figure not ideal enough, except breakage of grinding away, papery in image due to holding time length, preservation condition
Outside aging, there is also the coloring halo of many colours contaminate, the problem such as fade, even with not expecting that field color is deep or light to differ on portion drawing.And
Due to illumination effect during shooting, the colouring information of subregion weakens or loses original feature so that problem of extracting becomes
Obtain increasingly complex.
The analysis of threshold method based on color histogram that institute before uses, has merged gradient information and has believed with color
The problems referred to above can preferably be solved, its result such as Fig. 4 a and Fig. 4 b to Fig. 1 and Fig. 2 by breath.It can be seen that image
Extraction result preferable, big enable completes the identification to object and extraction, but the hydrographic curve sometimes extracted can exist broken string,
Belong to special case, at the broken string in Fig. 4 b.To carrying out careful observation at these broken strings, find that the main cause of broken string is Tu Zhongshui
Literary composition curve the thinnest so that have with grid lines overlap in a large number time (overlapping when usually curve is with grid lines less parallel) curve
Mass colour the most thoroughly cover the color of grid lines, the color of final display is by for the two superposition.The superposition of this color is led
Having caused the migration of colouring information, the pixel at this no longer meets universal hydrographic curve color characteristic, during extraction curve in this place
It is easy for producing broken string.What is more, when a certain section of curve with grid lines less parallel, it will produce substantial amounts of coincidence district
Territory, often breaks at this serious, it is difficult to carry out completion by methods such as expansions.Owing to no longer meeting threshold trait, original process
Method is the most applicable, and adjusts threshold value and noise spot will be caused to roll up, and image zooming-out is unstable.Then it is considered as new
Hydrographic curve is extracted by thinking.
Result after migrating in view of the colouring information of overlapping region not with curvilinear characteristic or grid lines feature etc.
Existing information is identical, and target is that the characteristic information after these being changed still is identified as aim curve.The present invention propose based on
The hydrographic curve extracting method of machine learning, by carrying out the sampling of multiple features fusion, it is thus achieved that some to pixel in image
The training sample of tape label, and utilize the method training of machine learning to obtain forecast model of classifying.Utilize the model can be to figure
Middle pixel carries out classification prediction, will be predicted as the pixel extraction of hydrographic curve out.And by classification error pixel region
Carrying out resampling, training sample set can be made to close the most complete, training produces the most healthy and the strongest classification forecast model.In order to not make
The influence of noise curve produced in processing procedure extracts result, the method for chain code following the to be utilized hydrographic curve to extracting
Image carries out post processing.
Concrete, the present invention comprises the steps of
Step A, determine window size and sampling feature combination, gather training sample;
First to determine the yardstick of sampling window when gathering training sample and need the feature of sampling.Sampling window is with currently
Pixel is window center point, takes into account in window the information of other pixels surrounding current pixel point, i.e. local message simultaneously.By
The information having in single pixel is fairly limited, when other local messages around pixel account for together, permissible
Make sample dimension higher, be more beneficial for careful classification.Window can use multiple yardstick, such as 3*3,5*5,7*7 etc., yardstick
The biggest, the information comprised in sample is the most, contributes to the most careful classifying accurately, but the time of calculating is the longest;Yardstick is the least then
Information in sample is the fewest, corresponding, calculates the time the shortest.Demand that concrete scale selection should be applied by reality and
Fixed.
On the other hand, the local message used during to window sample is also required to select, and these information are special by some
Value indicative forms, including color characteristic (RGB, Lab or HSI etc.), Gradient Features, textural characteristics (LBP or Gabor etc.), SIFT spy
Levy.The selection of concrete feature combination is considered as the impact on hydrographic curve extraction effect of these features, chooses the most appropriate
Several stack features be combined.Too much feature selection can bring the redundancy of information and the load of calculating, and very few feature then may be used
Classifying quality can be made to decline.The choice relation of feature combination effect and load when calculating of classification.
When one pixel is carried out above-mentioned sampling, should be according to the window size of agreement in advance, feature compound mode by certain
Order obtains each eigenvalue and is integrated into orderly sample vector.Gather color characteristic time, should according to from left to right, to
Under order successively each pixel in window is acquired.It addition, training sample is wanted other additional one-dimensional characteristic, as working as
The class label of front training sample.
When gathering training sample, choosing of sampled point is particularly important, it should however be noted that 1, should to take into account each when sampling different
Target classification, each should obtain enough sample points;2, in the pixel that target classification is identical, to contain as far as possible have
The pixel of different local features;3, identical in classification and pixel that local feature is similar should select several have typicality
Pixel is sampled.Wherein, the training sample set gathered should include at least " hydrographic curve ", " grid lines ", " other back ofs the body
Scape " sample of three kinds.
Step B, utilize machine learning method training produce classification forecast model;
Machine learning method includes the study of supervision and unsupervised study.Owing in current problem, class object is bright
Really, only want to extract hydrographic curve, therefore use the learning method having supervision, utilize the training sample of the band class label gathered
Obtain classification forecast model.This type of learning method includes decision tree, Bayes classifier, k nearest neighbor, BP neutral net, perceptron
And support vector machines etc..Different machine learning methods has different features, should select according to actual needs.Engineering
The choice relation of learning method effect and the efficiency that curve extracts.
Analyzing different hydrological data images to find, color and architectural feature between each image are much like, from feature
The angle in space, even different images substantially can also carry out classifying, extracting with the same component interface on feature space.
Then the method determining to use off-line learning, target is the classification forecast model that training produces an excellent effect, for institute
The pixel in pending image is had to carry out classifying, extracting rather than produce a model for the training of each image.
Step C, pending image is carried out classification prediction, and supplementary training sample set;
According to the mode arranged in above-mentioned steps A, the extraction characteristic of correspondence sample to pending image pixel by pixel,
And it is predicted classification as input via the classification forecast model obtained.Extract the picture being wherein predicted as " hydrographic curve "
Element, the result extracted as this curve.
If it is complete that curve extracts result, effect is satisfactory, can enter next step;But it is generally not capable of immediately obtained
Extracting result satisfactorily, the curve of extraction is the most relatively rough and there will be broken string, also can extract many noise spots.Solve
Certainly way is, constantly obtains new sample with incremental mode and adds training sample set, thus training is further improved and is good for
Strong classification forecast model.New training sample as increment both is from previous prediction classification results every time, the most therefrom looks for
Go out at curve broken string and other sort out the region that error rate is bigger, there is in selecting region on local feature the picture of typicality
Vegetarian refreshments is sampled.The method is intended to that past error prediction is carried out study and makes up, resampling near classification separating surface, by
This compensates the omission on sample space and vacancy, obtains the most complete training set, thus obtains and the finest demarcate accurately
Face and the most healthy and the strongest classification forecast model.
Utilize and add the classification forecast model of the training sample set re-training after increment and this image carried out again point
Class is predicted, if curve extraction effect is satisfactory, then this image is by currently processed, enters next step;Otherwise repeat above-mentioned
Increment adds the process of training sample.
The training process of classification forecast model is not stranghtforward, needs repeatedly to revise the new sample of increase and instructs
Practice;It is not to complete in some " training stage " made a clear distinction between good and evil simultaneously, but the best at the classifying quality of certain image
Time just start " retraining ";I.e., it has no one explicit and limited " training stage ".It addition, to training sample set
Increase the manually-operated intervention of needs, by the manual selected sampled point newly increased.But the initial accumulated rank due to training sample
Duan Tongchang can be quickly completed, and obtains the preferable model of effect, and only the most just may require that the existing mould of retraining
Type, the most manually-operated workload is the least.
Step D, chain code following carry out post processing.
Owing to original shooting image often exists much noise point, the curve of above-mentioned steps extracts still to exist in result to be permitted
Being difficult to the environment noise eliminated, they are that misalignment causes classification error to introduce mostly more.Due to the main pin of the present invention
Extraction to thinner hydrographic curve image, and remove noise if, with conventional corrosion or filtering method, often
Curve becomes very carefully the most seriously to break, and can not obtain satisfactory result.Preferably target is to be gone by noise spot
Remove, and hydrographic curve does not occur any change, post processing can be carried out in the way of taking chain code following in order to reach this target.
The information of each connected domain in figure can be followed the tracks of and be recorded to chain code following method in chain code mode, is each pixel
Labelling connected domain belonging to it, and record size and the bezel locations of each connected domain.The size of connected domain is not with wherein pixel
Number is as the criterion, and is as the criterion with the area of its minimum enclosed rectangle.
Result images after extracting classification, follows the tracks of the pixel being wherein predicted as " aim curve ", it is thus achieved that its connected domain
Information;That is, image is carried out the binaryzation of " aim curve/non-targeted curve ", and it is carried out above-mentioned chain code following.Its with
Track result will include real hydrographic curve target area and noise spot region, and the connected domain at the former place is the biggest, and
The latter is the least.Then the size of connected domain can be arranged specific threshold, thus get rid of those less, noise institutes
Connected domain.
It is so the most directly obtaining the connected domain of maximum, is for curve in the image of gained after preventing category of model from extracting
Yet suffer from broken string.This broken string is the trickleest, easily solves, and additionally can enter aim curve after removing noise spot
Row expansive working several times.
In order to improve treatment effect, it is also possible to cross Cheng Qian advanced person's row once to image copy at above-mentioned " curve tracing "
" grid lines tracking ", to determine grid lines region, and carries out above-mentioned " curve tracing " in this region.That is, secondary to image
Originally carry out the binaryzation of " grid lines/non-grid line ", and it carried out chain code following, obtain the bezel locations in largest connected territory,
Border line as grid lines.The purpose of this operation is to remove the outer all pixels unrelated with curve extraction of grid lines, reduces
Complexity during curve tracing.
In order to verify the effect of the present invention, choose several colored hydrology source map pictures and test, it is carried out above-mentioned
Classification extraction process.The selected window size size of agreement is 7*7, and the feature of institute's sampled point is combined as the RGB of each pixel
Color value, HSI color value and Lab color value 9 eigenvalues altogether;That is, the dimension of involved sample is 7*7*9=
441.And the machine learning method selected by arranging is support vector machines.As a example by Fig. 2, time initial, training sample set is combined into
Sky, first carries out initial samples to Fig. 2, it is thus achieved that after enough training samples, training generates SVM classifier, and for Fig. 2 is entered
Row classification prediction, its result such as Fig. 5 a, wherein black color dots is the point into " hydrographic curve " that predicts the outcome, and Grey Point is " grid
Line ".Visible, now to the classifying quality of Fig. 2 unsatisfactory, exist at many broken strings, especially occur in that two bigger
Broken position.To resampling at these broken strings several times, through several take turns re-training after generate SVM classifier effect carried
Height, all has been resolved at the classification results interrupt line of Fig. 2, such as Fig. 5 b.
Utilizing current SVM classifier to attempt Fig. 3 and carry out classification prediction, result such as Fig. 6 a, the curve now obtained is not deposited
In broken string phenomenon, but the most too many noise, classifying quality is the best.Again these noise spots are sampled, a training new round
Grader, again the result such as Fig. 6 b to Fig. 3 classification.Now classifying quality is preferable, it is believed that it is right that current class device can complete
The classificating requirement of these two figures.If it is desired, also other images can repeated aforesaid operations.
The classification results of Fig. 5 b Yu Fig. 6 b is proceeded post processing, removes unwanted " grid lines " Grey Point, utilize
The mode cancelling noise point of chain code following, and additionally curve is carried out expansive working several times, obtain curve and extract result such as Fig. 7 a
With Fig. 7 b.Visible, the present invention extraction to completing hydrographic curve thinner in Fig. 2 and Fig. 3.
The hydrographic curve extracting method based on machine learning method of the present invention, based on the classification to sample mode, as long as
Select sufficient training sample, be not required to consider the impact of the problems such as illumination;Use off-line learning, be not required to each figure
As Resurvey sample training;Select with incremental mode and add training sample, adapting to the new classificating requirement constantly arrived.
The present invention can preferably solve the disconnection problem easily produced when extracting fine rule, has good researching value.
Claims (10)
1. a hydrographic curve extracting method based on machine learning, it is characterised in that comprise the following steps:
Step A, the yardstick of selected sampling window and the target characteristic that need to sample, and gather representative training sample accordingly
Set;
Step B, utilize machine learning method train from training sample generation classification forecast model;
Step C, to each pixel in pending image, collect target characteristic as sample to be sorted according to sampling window
This, utilize described classification forecast model to classify;
Step D, judge in pending image whether the classification results of each pixel reaches expection, curve whether extract complete and
Whether there is the obvious region of other classification errors;If classification results reaches expection, then enter step F;Otherwise, step E is entered;
Step E, from curve broken string region and the obvious region of classification error choose representative sample point, to its sample
After add in training sample set, and repeat step B;
Step F, to process after image carry out post processing.
2. hydrographic curve extracting method based on machine learning as claimed in claim 1, it is characterised in that in step A, described window
The yardstick of mouth is scalable.
3. hydrographic curve extracting method based on machine learning as claimed in claim 1, it is characterised in that in step A, to local
The combination selecting the various types of feature of employing of feature, described feature includes color characteristic, Gradient Features, textural characteristics
And SIFT feature.
4. hydrographic curve extracting method based on machine learning as claimed in claim 1, it is characterised in that in step B, described machine
Device learning method includes support vector machine, neural net method and combinations thereof.
5. hydrographic curve extracting method based on machine learning as claimed in claim 1, it is characterised in that in step C, travel through institute
There is the pixel that can extract feature with current sampling window, obtain each corresponding local feature value forming according to window and treat point
Class sample vector.
6. hydrographic curve extracting method based on machine learning as claimed in claim 1, it is characterised in that in step D, for dividing
Class result whether preferably decision factor includes that curve is the most complete, noise whether limited amount and be not result in aim curve
Misjudgement.
7. hydrographic curve extracting method based on machine learning as claimed in claim 1, it is characterised in that step E is further:
The sample point chosen should add in original training set with incremental form as training sample and re-training produces prediction mould
Type, to carry out directive adjustment to original model.
8. as claimed in claim 1 hydrographic curve extracting method based on machine learning, it is characterised in that in step F, described after
Processing mode is that chain code following combines with expansion process.
9. hydrographic curve extracting method based on machine learning as described in claim 1 or 6, it is characterised in that also need to add
Sample class " grid lines ", in order to positioning as graph region on figure during process image.
10. hydrographic curve extracting method based on machine learning as claimed in claim 8, it is characterised in that described chain code following
Connected domain present in the tracking image, calculate, record the size of these connected domains, thus gets rid of and is made up of noise
The connected domain that size is less.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610520993.4A CN106203496B (en) | 2016-07-01 | 2016-07-01 | Hydrographic curve extracting method based on machine learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610520993.4A CN106203496B (en) | 2016-07-01 | 2016-07-01 | Hydrographic curve extracting method based on machine learning |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106203496A true CN106203496A (en) | 2016-12-07 |
CN106203496B CN106203496B (en) | 2019-07-12 |
Family
ID=57466160
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610520993.4A Active CN106203496B (en) | 2016-07-01 | 2016-07-01 | Hydrographic curve extracting method based on machine learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106203496B (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108090686A (en) * | 2017-12-29 | 2018-05-29 | 北京大学 | A kind of medical events risk-assessment method and system |
CN109753971A (en) * | 2017-11-06 | 2019-05-14 | 阿里巴巴集团控股有限公司 | Distort the antidote and device, character identifying method and device of literal line |
CN111539587A (en) * | 2020-03-06 | 2020-08-14 | 李�杰 | Hydrological forecasting method |
CN113359216A (en) * | 2021-06-03 | 2021-09-07 | 山东捷瑞数字科技股份有限公司 | Method, system and storage medium for identification of recorded data of tipping-bucket rain gauge |
CN114648570A (en) * | 2022-03-28 | 2022-06-21 | 杭州电子科技大学 | Curve extraction method for differential background grids based on deep learning |
CN115409825A (en) * | 2022-09-06 | 2022-11-29 | 重庆众仁科技有限公司 | Temperature-humidity pressure trace identification method based on image identification |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20120070044A1 (en) * | 2010-09-21 | 2012-03-22 | General Electric Company | System and method for analyzing and visualizing local clinical features |
CN103020971A (en) * | 2012-12-28 | 2013-04-03 | 青岛爱维互动信息技术有限公司 | Method for automatically segmenting target objects from images |
CN103971367A (en) * | 2014-04-28 | 2014-08-06 | 河海大学 | Hydrologic data image segmenting method |
CN104573731A (en) * | 2015-02-06 | 2015-04-29 | 厦门大学 | Rapid target detection method based on convolutional neural network |
US9053551B2 (en) * | 2012-05-23 | 2015-06-09 | International Business Machines Corporation | Vessel identification using shape and motion mapping for coronary angiogram sequences |
CN105184265A (en) * | 2015-09-14 | 2015-12-23 | 哈尔滨工业大学 | Self-learning-based handwritten form numeric character string rapid recognition method |
-
2016
- 2016-07-01 CN CN201610520993.4A patent/CN106203496B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20120070044A1 (en) * | 2010-09-21 | 2012-03-22 | General Electric Company | System and method for analyzing and visualizing local clinical features |
US9053551B2 (en) * | 2012-05-23 | 2015-06-09 | International Business Machines Corporation | Vessel identification using shape and motion mapping for coronary angiogram sequences |
CN103020971A (en) * | 2012-12-28 | 2013-04-03 | 青岛爱维互动信息技术有限公司 | Method for automatically segmenting target objects from images |
CN103971367A (en) * | 2014-04-28 | 2014-08-06 | 河海大学 | Hydrologic data image segmenting method |
CN104573731A (en) * | 2015-02-06 | 2015-04-29 | 厦门大学 | Rapid target detection method based on convolutional neural network |
CN105184265A (en) * | 2015-09-14 | 2015-12-23 | 哈尔滨工业大学 | Self-learning-based handwritten form numeric character string rapid recognition method |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109753971A (en) * | 2017-11-06 | 2019-05-14 | 阿里巴巴集团控股有限公司 | Distort the antidote and device, character identifying method and device of literal line |
CN109753971B (en) * | 2017-11-06 | 2023-04-28 | 阿里巴巴集团控股有限公司 | Correction method and device for distorted text lines, character recognition method and device |
CN108090686A (en) * | 2017-12-29 | 2018-05-29 | 北京大学 | A kind of medical events risk-assessment method and system |
CN108090686B (en) * | 2017-12-29 | 2022-01-25 | 北京大学 | Medical event risk assessment analysis method and system |
CN111539587A (en) * | 2020-03-06 | 2020-08-14 | 李�杰 | Hydrological forecasting method |
CN111539587B (en) * | 2020-03-06 | 2023-11-24 | 武汉极善信息技术有限公司 | Hydrologic forecasting method |
CN113359216A (en) * | 2021-06-03 | 2021-09-07 | 山东捷瑞数字科技股份有限公司 | Method, system and storage medium for identification of recorded data of tipping-bucket rain gauge |
CN114648570A (en) * | 2022-03-28 | 2022-06-21 | 杭州电子科技大学 | Curve extraction method for differential background grids based on deep learning |
CN114648570B (en) * | 2022-03-28 | 2024-03-26 | 杭州电子科技大学 | Curve extraction method for differentiated background grid based on deep learning |
CN115409825A (en) * | 2022-09-06 | 2022-11-29 | 重庆众仁科技有限公司 | Temperature-humidity pressure trace identification method based on image identification |
CN115409825B (en) * | 2022-09-06 | 2023-09-12 | 重庆众仁科技有限公司 | Temperature, humidity and pressure trace identification method based on image identification |
Also Published As
Publication number | Publication date |
---|---|
CN106203496B (en) | 2019-07-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106203496A (en) | Hydrographic curve extracting method based on machine learning | |
CN107871124B (en) | A kind of Remote Sensing Target detection method based on deep neural network | |
CN109284669A (en) | Pedestrian detection method based on Mask RCNN | |
CN109583425A (en) | A kind of integrated recognition methods of the remote sensing images ship based on deep learning | |
CN109902806A (en) | Method is determined based on the noise image object boundary frame of convolutional neural networks | |
CN103049763B (en) | Context-constraint-based target identification method | |
CN112036335B (en) | Inverse convolution guided semi-supervised plant leaf disease identification and segmentation method | |
WO2018107939A1 (en) | Edge completeness-based optimal identification method for image segmentation | |
US20180053033A1 (en) | Applying Pixelwise Descriptors to a Target Image that are Generated by Segmenting Objects in Other Images | |
CN108875595A (en) | A kind of Driving Scene object detection method merged based on deep learning and multilayer feature | |
CN110135296A (en) | Airfield runway FOD detection method based on convolutional neural networks | |
CN104408449B (en) | Intelligent mobile terminal scene literal processing method | |
CN110532946B (en) | Method for identifying axle type of green-traffic vehicle based on convolutional neural network | |
CN105809121A (en) | Multi-characteristic synergic traffic sign detection and identification method | |
CN109214420A (en) | The high texture image classification method and system of view-based access control model conspicuousness detection | |
CN112686902B (en) | Two-stage calculation method for brain glioma identification and segmentation in nuclear magnetic resonance image | |
CN110992381A (en) | Moving target background segmentation method based on improved Vibe + algorithm | |
CN110008900B (en) | Method for extracting candidate target from visible light remote sensing image from region to target | |
CN110059760A (en) | Geometric figure recognition methods based on topological structure and CNN | |
CN106529532A (en) | License plate identification system based on integral feature channels and gray projection | |
CN104657980A (en) | Improved multi-channel image partitioning algorithm based on Meanshift | |
CN104680193B (en) | Online objective classification method and system based on quick similitude network integration algorithm | |
CN109766823A (en) | A kind of high-definition remote sensing ship detecting method based on deep layer convolutional neural networks | |
CN110008899B (en) | Method for extracting and classifying candidate targets of visible light remote sensing image | |
CN107016680B (en) | A kind of pest image background minimizing technology based on conspicuousness detection |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |