CN110533971A - A kind of intelligent tutoring system deeply interacted - Google Patents
A kind of intelligent tutoring system deeply interacted Download PDFInfo
- Publication number
- CN110533971A CN110533971A CN201910655383.9A CN201910655383A CN110533971A CN 110533971 A CN110533971 A CN 110533971A CN 201910655383 A CN201910655383 A CN 201910655383A CN 110533971 A CN110533971 A CN 110533971A
- Authority
- CN
- China
- Prior art keywords
- information
- teaching
- student
- unit
- interacted
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/30—Authentication, i.e. establishing the identity or authorisation of security principals
- G06F21/31—User authentication
- G06F21/32—User authentication using biometric data, e.g. fingerprints, iris scans or voiceprints
-
- G—PHYSICS
- G09—EDUCATION; CRYPTOGRAPHY; DISPLAY; ADVERTISING; SEALS
- G09B—EDUCATIONAL OR DEMONSTRATION APPLIANCES; APPLIANCES FOR TEACHING, OR COMMUNICATING WITH, THE BLIND, DEAF OR MUTE; MODELS; PLANETARIA; GLOBES; MAPS; DIAGRAMS
- G09B5/00—Electrically-operated educational appliances
- G09B5/08—Electrically-operated educational appliances providing for individual presentation of information to a plurality of student stations
- G09B5/14—Electrically-operated educational appliances providing for individual presentation of information to a plurality of student stations with provision for individual teacher-student communication
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Security & Cryptography (AREA)
- Business, Economics & Management (AREA)
- Educational Administration (AREA)
- Educational Technology (AREA)
- Computer Hardware Design (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The present invention provides a kind of intelligent tutoring systems deeply interacted, comprising: teaching end for sending interactive teaching and learning information to study end, and is inquired the authentication information and return information at study end；Learn end, for realizing the biometric information authentication of student, and receives and reply the interactive information that teaching end is sent；Storage end, for storing teaching end and learning the data information of end generation.The present invention provides a kind of intelligent tutoring system deeply interacted, the time of Faculty and Students' interaction is greatly shortened by technical approach, allows the thinking and idea of each available student of teacher, to promote quality of instruction.
Description
Technical field
The present invention relates to intelligent tutoring technical field more particularly to a kind of intelligent tutoring systems deeply interacted.
Background technique
Interactive teaching is exactly the teaching environment by building multilateral interaction, in the process of teaching both sides' equality communication and discussion
In, reach different viewpoints collision blending, and then excite the initiative and exploration of teaching both sides, reaches and improve the one of teaching efficiency
Kind teaching method.This method theme is clear, and orderliness understands, inquires into deeply, can give full play to enthusiasm, the creativeness of student.
Currently, being normally based on the discussion of some problems during implementing interactive teaching, being just able to achieve exchange
Purpose increases the degree of participation of student to promote classroom atmosphere.But the discussion for problem, it does not ensure that whole
Student can put into wherein, and especially in modern university campus, as soon as common course is all that several classes piece carry out, this is big
The popularization and development of interactive teaching are limited greatly, is based on disadvantages described above, interactive teaching is applied only for class student now
The elective of negligible amounts, for compulsory course, teacher often selects to be unable to reach optimal using common teaching model
Teaching efficiency.
Summary of the invention
The present invention in view of the above problems, in order to overcome interactive teaching can not be applied to extensive classroom lack
It falls into, the present invention provides a kind of intelligent tutoring systems deeply interacted, greatly shorten Faculty and Students' interaction by technical approach
Time, the thinking and idea of each available student of teacher are allowed, to promote quality of instruction.
The present invention is realized by following measure:
A kind of intelligent tutoring system deeply interacted, comprising:
Impart knowledge to students end, for study end send interactive teaching and learning information, and to study end authentication information and return information into
Row inquiry；
Learn end, for realizing the biometric information authentication of student, and receives and reply the interactive information that teaching end is sent；
Storage end, for storing teaching end and learning the data information of end generation.
Further, the teaching end includes:
Information issuing module for uploading interactive teaching and learning information, and sends this information to each study end；
Enquiry module, authentication information and return information for query learning end.
Further, the study end includes:
Authentication module carries out typing and identification certification for biological informations such as fingerprint, faces to each student；
Interactive module, teaching interaction information and simultaneous display for receiving teaching end are handed in the touch screen at study end
Mutual information realizes presentation process according to the guide of demonstration instruction by the way of human assistance demonstration, by student.
Further, the study end further include:
Course recommending module by the similarity of all previous student's electives of calculating, or analyzes each elective
Weight, auxiliary student carry out the selection of elective course.
Further, the course recommending module includes:
Similarity analysis unit, the unit are based on multiple view hash algorithm, carry out to the historical data of all previous students
Baseline encoded is obtained after coding, real-time coding is then calculated according to the real time data of curricula-variable student, by comparing real-time coding
And baseline encoded, it finds and the highest historical data of real time data similarity；Or
Weight analysis unit is based on comprehensive weight analytic approach, analyzes weight of each elective compared to specialized course
Coefficient, discovery influence maximum elective course to specialized course.
Further, the study end further include:
It is bright to adjust screen for the corresponding relationship of illuminance and screen intensity by prestoring in real time for brightness coarse adjustment module
Degree.
Further, the brightness coarse adjustment module includes:
First setup unit, for setting the corresponding relationship of illuminance and screen intensity；
Illuminance measurement unit, for detecting the illuminance in the study end external world；
Unit, the illuminance for obtaining according to illuminance measurement unit are adjusted, corresponding according to setup unit setting is closed
System adjusts screen intensity.
Further, the study end further include:
Brightness accurate adjustment module, for adjusting screen intensity according to the relative positional relationship of human eye and screen.
Further, the brightness accurate adjustment module includes:
Second setup unit, for setting the corresponding relationship between human eye distance, angle and screen intensity three；
First position determination unit, for obtaining face relative to using binocular restructing algorithm on the basis of learning end
Practise position and the posture at end；
Second position determination unit, for obtaining human eye relative to using the cascade classifier integrated in OpenCV
Practise the position at end；
The third place determination unit calculates human eye and study for the relative positional relationship according to study end and human eye
Hold the distance and angle of screen；
Accurate adjustment unit, for adjusting according to the calculated result of the third place unit and the corresponding relationship of the second setup unit
Screen intensity.
Technical solution of the present invention has the beneficial effect that
For in the prior art using problem discussion as the interactive teaching method of core, the application is that student is equipped with study
End is equipped with teaching end for teacher, teaching demonstration and problem discussion is spread to each student, a side by technical approach
Face, teacher and each student are connected, are capable of forming and are similar to one-to-one Mutual Teaching Pattern by which, another
Aspect, teacher can know the study situation and the mode of thinking of the student, more specific aim by the feedback information of each student
Promotion quality of instruction and teaching link.
It is both selection of the teacher to the selection and student of student to teacher, especially in university in interactive teaching
Elective course time of student and teacher can be only wasted, therefore, for students' needs if having selected inappropriate elective course
Link student can be assisted to select elective course interested or useful by similarity analysis or weight analysis, help student
Obtain best learning effect.
In addition, study end, as an electronic curtain, if screen intensity is not suitable for, the eyesight influence to student is very big
, therefore, uses extraneous illuminance and the associated method of screen intensity, realize the adjusting of screen intensity.Even more into one
Step, using the association of human eye distance, angle and screen, realize that the screen intensity more refined is adjusted, further protection
Raw eyesight.
Detailed description of the invention
Fig. 1 is the schematic illustration of present system.
Specific embodiment
In order to clarify the technical characteristics of the invention, below by specific embodiment, and its attached drawing is combined, to this hair
It is bright to be described in detail.Following disclosure provides many different embodiments or example is used to realize different knots of the invention
Structure.In order to simplify disclosure of the invention, hereinafter the component of specific examples and setting are described.In addition, the present invention can be with
Repeat reference numerals and/or letter in different examples.This repetition is that for purposes of simplicity and clarity, itself is not indicated
Relationship between various embodiments and/or setting is discussed.It should be noted that illustrated component is not necessarily to scale in the accompanying drawings
It draws.Present invention omits the descriptions to known assemblies and treatment technology and process to avoid the present invention is unnecessarily limiting.
A kind of intelligent tutoring system deeply interacted, the system include teaching end, study end and storage end.
The teaching end, for sending interactive teaching and learning information to study end, and authentication information and reply to study end
Information is inquired.Wherein, teaching end includes an information issuing module, for uploading interactive teaching and learning information, and by the information
It is sent to each study end, the information issued in the function includes but is not limited to that video, picture, audio, table etc. are different
Data format or presentation mode, the information of publication have the effect of playing and manual secondary presentation automatically.In addition, teaching end is also wrapped
An enquiry module is included, for the authentication information and return information at query learning end, such as when what is discussed to the transmission of study end asks
When topic, the problem of learning end reply can be checked by the module.
The study end as the main interaction platform of the application, from allomeric function for for realizing student life
Object authentification of message, and the interactive information that teaching end is sent is received and replys, for learning the concrete function and corresponding module at end,
Elaboration as detailed below.
Authentication module carries out typing and identification certification, the mould for biological informations such as fingerprint, faces to each student
The effect of block is on the one hand to realize function of registering, and saves the process of teacher's roll-call, and teacher can appointing in classroom
Meaning time point carries out authentication operation, convenient and efficient；On the other hand, which is the identity in order to identify student, so as to
The case where enough allowing teacher to learn each learner answering questions problem.
Interactive module, teaching interaction information and simultaneous display for receiving teaching end are handed in the touch screen at study end
Mutual information realizes presentation process according to the guide of demonstration instruction by the way of human assistance demonstration, by student.Demonstration instruction can
To allow student to participate in teaching process, realize better interactive experience using the prompt of text or voice.
Course recommending module, the course recommending module include similarity analysis unit or weight analysis unit, are led to
The similarity for calculating all previous student's electives is crossed, or the weight of each elective of analysis, auxiliary student carry out elective course
Selection.
Similarity analysis unit therein is based on multiple view hash algorithm, carries out to the historical data of all previous students
Baseline encoded is obtained after coding, real-time coding is then calculated according to the real time data of curricula-variable student, by comparing real-time coding
And baseline encoded, it finds and the highest historical data of real time data similarity.
In the algorithm, the specific implementation principle and process of baseline encoded are as follows:
Before system operation, obtains in advance and store historical data, analytical unit are utilized by a preprocessing process
This open source of Storm is handled in real time is filtered screening, data schema with computing technique, to pretreated data, uses
HBase storage obtains baseline encoded after encoding to all historical datas then by a cataloged procedure.
The detailed process of coding are as follows:
Input following parameter: hashcode digit k, number of views m, number of students n, student's similarityStudent characteristics vector
Combination algorithm HashingCodeLearning (k, m, n,), export following parameter: student's totality Hash is compiled
Code U, each view weight α, each view hash function
Initialization
Construct connection matrix
Construct Laplacian Matrix (Dp)-1/2LP(Dp)-1/2, p=1,2 ..., m judge whether to restrain, if not converged, follow
The following calculating process of ring:
It is calculated
It is calculated
Matrix is calculated
The smallest feature vector of k corresponding eigenvalue of matrix H (α) is calculated；
Hash encoder matrix U is generated according to feature vector；
It is calculated
α is obtained using Novel Algorithm；
U is returned,α；
After obtaining serialization Hash coding, binaryzation is carried out to it, obtains the benchmark that each value is -1 or 1
Coding.
For weight analysis unit, which uses subjective weighting method-analytic hierarchy process (AHP), carries out tax power to index；
With objective weighted model-entropy assessment, tax power is carried out to index；With comprehensive weight method, above-mentioned subjective, objective weight is carried out
COMPREHENSIVE CALCULATING obtains new weighted value, analyzes weight coefficient of each elective compared to specialized course, finds to specialized course
Influence maximum elective course.
The specific implementation principle of comprehensive weight analytic approach are as follows:
The Main Analysis step of analytic hierarchy process (AHP) is:
1, hierarchy Model is established.On the basis of analysing in depth practical problem, by related each factor according to not
Resolve into several levels from top to down with attribute, the factors of same layer are subordinated to one layer of factor or to upper layer because being known as
It influences, while dominating next layer of factor or the effect by lower layer factors again.Top layer is destination layer, usually only 1 because
Element, lowest level are usually scheme or object layer, and centre can have one or several levels, usually criterion or indicator layer.Work as standard
Sub- rule layer should further be decomposited by (being for example more than 9) when then excessive.
2, Paired comparison matrix is constructed.It is one layer upper for being subordinated to (or influence) since the 2nd layer of hierarchy Model
The same layer factors of each factor compare dimensional configurations Paired comparison matrix with Paired Comparisons and 1-9, until lowest level.
1) according to scaling theory, multilevel iudge matrix A two-by-two is constructed:
In formula, aij=1, aij=1/aji；
2) each column of judgment matrix A are made into normalized:
3) the sum of each row element of judgment matrix A is soughtCalculation formula are as follows:
4) rightIt is normalized to obtain wi, calculation formula are as follows:
5) according to Aw=λmaxW finds out Maximum characteristic root and its feature vector.
3, it calculates weight vector and does consistency check.Maximum characteristic root and corresponding spy are calculated for each Paired comparison matrix
Vector is levied, does consistency check using coincident indicator, random index and consistency ratio.If upchecking, feature
Vector (after normalization) is weight vector: if not passing through, need to reconfigure in pairs relatively battle array.
1. calculating coincident indicator
2. finding out corresponding Aver-age Random Consistency Index；
3. calculating consistency ration C.R.=C.I./R.I.；
As C.R. < 0.1, it is subjected to consistency check, otherwise A is corrected.
It calculates right vector and does combination consistency check.Lowest level is calculated to the right vector of target, and according to
Formula does combination consistency check, if upchecking, the result that can be indicated according to right vector carries out decision, otherwise needs
It rethinks model or reconfigures the biggish Paired comparison matrix of those consistency ratios.
The calculating step of entropy assessment are as follows:
Step 1: index is normalized, wherein the calculation formula of positive index are as follows: The higher the better for the numerical value；The calculation formula of negative sense index are as follows: The numerical value it is more low more；
Step 2: the specific gravity that i-th of sample value under jth item index accounts for the index is calculated, calculation formula are as follows:
Step 3: calculating the entropy of jth item index, calculation formula are as follows:Wherein k=
1/ln (i)；
Step 4: comentropy redundancy (difference) is calculated, calculation formula are as follows: dj=1-ej；
Step 5: calculating the weight of each index, calculation formula are as follows:
The main calculating step of comprehensive weight is:
1, comprehensively consider entropy assessment and analytic hierarchy process (AHP) obtain weight, calculation formula are as follows:
2, to normalized:
In addition to this, during the realization of weight analysis unit, in order to preferably realize weight analysis, in comprehensive weight
Before analysis, index dimensionality reduction is carried out using principal component analysis, guarantees simplifying for index system, when the index quantity of data is greater than 3
When a, with Principal Component Analysis, retain the index that contribution rate is greater than 90%；When the index quantity of data is less than or equal to 3,
Directly retain whole indexs, finally obtains the index system of each dimension.
The calculating step of principal component analysis is:
1, the standardized acquisition of original index data, { xpOverall objective factor collection.
2, standardized transformation,
NoteQ introduces new parameter α=Q to standardize orthogonal matrix, and Z=XQTβ
Or β=Q α, then y=β01+ZQTβ+ε=β01+Zα+ε。
3, the characteristic equation for solving sample correlation matrix, obtains p characteristic root, determines principal component.
X at this timeTThe eigenvalue λ of XiI-th of principal component z is measurediThe size of value variation, i.e., each master in n times experiment
Ingredient is to weight coefficient shared in general impacts.If λi≈ 0 then shows to reject i-th of principal component；If preceding r principal component institute
The ratio accounted for close enough 100%, then subsequent principal component can also be rejected directly.Wherein, claimFor i-th principal component
Contribution rate.
4, the target variable after standardization is converted into principal component, Uij=zi Tbj, j=1,2 ..., m, U1Referred to as first is main
Ingredient, U2Referred to as Second principal component, ..., UpReferred to as pth principal component.
5, overall merit is carried out to m principal component.Summation is weighted to get final evaluation of estimate, flexible strategy to m principal component
For the variance contribution ratio of each principal component.
It is bright to adjust screen for the corresponding relationship of illuminance and screen intensity by prestoring in real time for brightness coarse adjustment module
Degree.The module includes following part:
First setup unit, for setting the corresponding relationship of illuminance and screen intensity, the corresponding relationship is with data form
Form be stored in storage end.
Illuminance measurement unit, using one or more illuminance sensors, for detecting the illuminance in the study end external world.
Adjust unit, the illuminance for obtaining according to illuminance measurement unit, in conjunction with pair of illuminance and screen intensity
Relation table is answered, screen intensity is adjusted.
For the adjusting screen intensity more refined, it is also provided with a brightness accurate adjustment module, for according to human eye and screen
The relative positional relationship of curtain adjusts screen intensity.
The brightness accurate adjustment module includes:
Second setup unit, for setting the corresponding relationship between human eye distance, angle and screen intensity three.
First position determination unit, for obtaining face relative to using binocular restructing algorithm on the basis of learning end
Practise position and the posture at end.The realization principle of the unit are as follows:
Binocular restructing algorithm is the data image that object scene is obtained by video camera, and is carried out at analysis to this image
Reason, the method for deriving the three-dimensional information of object in actual environment in conjunction with computer vision knowledge.
Can mainly have these steps: image obtains, camera calibration obtains ginseng outside internal reference, feature extraction, three-dimensional
Match, three-dimensional building.
It is the pictorial information for teaching building object different angle to difference based on common RGB camera that image, which obtains,.
It is a necessary step that camera calibration, which obtains ginseng outside internal reference, because the video camera of different model throws real-world object
The conversion parameter of shadow imaging is different, the also just difference of transition matrix when doing picture reverse transformation to three-dimensional reconstruction.
Join outside video camera: determining relative positional relationship between camera coordinates and world coordinate system.
Video camera internal reference: projection relation of the video camera from three-dimensional space to two dimensional image is determined.
If P=(X, Y, Z) P=(X, Y, Z) is a bit in scene, in pinhole camera model, to pass through following several
A transformation eventually becomes picture point p=(μ, ν) p=(μ, ν) on two dimensional image:
P is transformed into camera coordinates system by rigid body translation (rotation and translation) from world coordinate system, this conversion process
Use the relative pose between camera, that is, the outer parameter of camera.
From camera coordinates system, pass through the picture point p=(x, y) on the imaging plane of perspective projection transformation to camera.
By picture point pp from imaging coordinate system, pixel coordinate is transformed to by zooming and panning and fastens point p=(μ, ν).
Three-dimensional point in scene is transformed to the two-dimensional points in image, that is, the combination of each coordinate system transformation by camera,
Conversion process above can be arranged to the form for matrix multiple:
Matrix K is known as to the intrinsic parameter of camera,
Wherein, α, β indicate the number of pixel in unit distance on image, then fx=α f, and fy=β f becomes the focal length f of camera
It is changed to the pixel measurement representation on x, the direction y.
In addition, can add a warp parameters γ on the internal reference matrix of camera in order to without loss of generality, which is used
To indicate the distortion of two reference axis of pixel coordinate system.Then intrinsic parameter K becomes
More famous scaling method --- Zhang Shi standardization, it both overcomes the high-precision three-dimensional that photography standardization needs
The shortcomings that demarcating object, and solve the problems, such as self-calibration method poor robustness.We are with chessboard as the calibration object of camera calibration
(object in digital picture is mapped to from real world), chessboard are one piece of scaling boards being made of black and white party block gap.Institute
It uses chessboard as calibration object using us and is because plane checker board pattern is easier to handle (relative to complicated three-dimension object), but with
This simultaneously, two-dimensional bodies can lack a part of information relative to three-dimension object, then we can be varied multiple times chessboard orientation come
Image is captured, in the hope of obtaining richer coordinate information.
It is aware of the coordinate of the corresponding points of two planes, so that it may which solution obtains the homography matrix H of two planes.Wherein,
The gridiron pattern of calibration be it is special, the coordinate of angle point is known；Angle point in image can be obtained by Robust Algorithm of Image Corner Extraction
It arrives, can be obtained by the homography matrix H of chessboard plane Π and plane of delineation π in this way.
Have by camera model above: p=K [R | t] P.Wherein, p is picpointed coordinate, and P is the chessboard coordinate of calibration.This
Sample can be obtained by following equation: and H=K [R | t], what H was indicated is singly to answer square between imaging plane and calibration chessboard plane
Battle array.By corresponding point to H is solved after, then the intrinsic parameter K of camera can be obtained by above equation, and join spin moment outside
Battle array R and translation vector t.
Feature mainly includes characteristic point, characteristic curve and region.It is all in most cases the spy using characteristic point as Matching unit
Sign point extracts in what manner to be closely connected with which kind of matching strategy.Therefore it needs first to determine when carrying out the extraction of characteristic point
With which kind of matching process.The corresponding points in two images are needed, the extraction of this characteristic point reformed into and matching problem.For figure
The not larger situation of aberration, it is recommended to use SIFT (Scale invariant features transform) algorithm, because SIFT is to rotation, scale, perspective
There is preferable robustness.If difference is little, it may be considered that other faster features, such as SURF, ORB etc., there are also one
Kind is based on the improved SURF of SIFT (rapid robust feature) algorithm.
SIFT has 4 key steps: 1) extremum extracting of scale space is searched for the image on all scale spaces, is passed through
Gaussian derivative function potentially to scale and selects constant point of interest to identify.2) positioning feature point, in the position of each candidate
It sets, Location Scale, their degree of stability of the basis for selecting of key point is determined by a fitting refined model.3) feature
Direction assignment, the gradient direction based on image local distribute to each key point position one or more direction, subsequent all
Operation is all that the direction for key point, scale and position convert, to provide the invariance of these features.4) characteristic point
Description measures the partial gradient of image in the neighborhood around each characteristic point on selected scale, these gradients are transformed
At a kind of expression, this deformation and the light change for indicating to allow bigger local shape.
Gaussian kernel is the core that uniquely can produce multiscale space.
The scale space L (x, y, σ) of one image, is defined as original image I (x, y) and 2 dimensions of a variable dimension are high
This function G (x, y, σ) convolution algorithm.
Two-dimensional space Gaussian function expression formula:
The scale space of image is exactly: after two-dimensional Gaussian function and original image convolution algorithm as a result, scale space
Expression formula:
L (x, y, σ)=G (x, y, σ) * I (x, y)
Gaussian kernel is that circle is symmetrical, and what is shown in picture pixels is a square, and size is by Gaussian template
It determines.The result of convolution makes original pixel value have maximum weight, and the remoter adjacent pixel values weight of distance center is also smaller.
" scale-space representation " refer to different Gaussian kernels smoothed out picture different expression, the meaning is exactly: point of original photo
Resolution, with by the smoothed out photo of different Gaussian kernels resolution ratio be as.But for computer, different moulds
Paste degree, the appearance that photo " seeing " gets on is with regard to different.Gaussian kernel is bigger, and picture " seeing " gets on fuzzyyer.Computer does not have
Subjective consciousness goes where identification is characteristic point, what it can do, only tells the most fast point of change rate.Cromogram is triple channel
, bad detection catastrophe point.Need RGB figure being converted to grayscale image, at this time grayscale image be single channel, gray value 0~255 it
Between be distributed.
No matter human eye observation's photo distance it is how far, as long as the contour feature of object key can be recognized, that
Substantially know information expressed by image.Computer is same, and after Gaussian convolution, although image fogs.But it is whole
Pixel do not become, can still find the point of gray scale value mutation.And these points, so that it may as candidate feature point, later period
Quantity a little is further reduced, accuracy rate is improved.
In the multiple dimensioned usually used image pyramid representation of earlier picture.Image pyramid is same image not
One group obtained under same resolution ratio is as a result, its generating process generally comprises two steps: 1) carrying out to original image smooth；2)
Down-sampled (usually horizontal, vertical direction 1/2) is carried out to treated image.
A series of image of continuous size reductions is obtained after down-sampled.Obviously, in a traditional pyramid, each layer
Image is that a tomographic image is long thereon, high each half.Although the image pyramid of multiresolution generates simply, its essence is drop
Sampling, the local feature of image is then difficult to keep, that is, cannot keep the scale invariability of feature.
After gaussian pyramid constructs successfully, each group, which is subtracted each other for adjacent two layers, can be obtained by DoG pyramid
In order to find the extreme point of scale space, each pixel will be with its image area (same scale space) and scale
All consecutive points in domain (adjacent scale space) are compared, and when it is greater than (or being less than) all consecutive points, are changed the time just
It is extreme point.As shown, intermediate test point will be with 3 × 33 × 3 neighborhood, 8 pixels and its phase of image where it
Adjacent 18, bilevel 3 × 33 × 3 field pixel, totally 26 pixels are compared.
It is recognised that the first layer of every group of image and the last layer are can not be compared and obtain pole in from the description above
Value.In order to meet the continuity of change of scale, Gaussian Blur is continued to use in the top layer of each group of image and generates 3 width images,
Every group of gaussian pyramid has S+3S+3 tomographic image, and pyramidal every group of DoG has S+2S+2 group image.
And the extreme point that we look for is decided after difference of Gaussian, then it is belonged on discrete space
Point, be not necessarily extreme point truly.
Discrete to be converted to continuously, we will recognize that Taylor expansion:
Then extreme point are as follows:
Some extreme points are not that we want, and just having most in the middle is the extreme point that fringe region generates.Because
The edge contour of object is in grayscale image, and there is the mutation of gray value, such point just " is takeed for " being special in calculating
Value indicative.
It carefully analyzes, gray scale value mutation is very big in the longitudinal direction for fringe region, but the variation in transverse direction is with regard to very little.
Due to this special nature, we have suggested Hessian matrix, Hesse matrices be for seeking curvature, can be with
The second order local derviation of function is element, constitutes the matrix H of a 2x2:
In order to realize image rotation invariance, the direction to characteristic point is needed to carry out assignment.Utilize characteristic point neighborhood territory pixel
Gradient distribution characteristic determine its directioin parameter, recycle the histogram of gradients of image to seek the stabilization of key point partial structurtes
Direction.
Characteristic point is had found, the scale σ of this feature point is also can be obtained by, also can be obtained by the scale where characteristic point
Image L (x, y)=G (x, y, σ) * I (x, y) is calculated centered on characteristic point, using 3 × 1.5 σ as the width of the area image of radius
Angle and amplitude, the mould m (x, y) and direction θ (x, y) of the gradient of each point L (x, y) can acquire gradient width table by following company
Up to formula:
Gradient direction expression formula:
Be calculated after gradient direction it is necessary to use in histogram statistical features vertex neighborhood the corresponding gradient direction of pixel and
Amplitude.The horizontal axis of the histogram of gradient direction is that (range of gradient direction is 0 to 360 degree, and histogram is every for the angle of gradient direction
36 degree of columns totally 10 columns, or do not have 45 degree of columns totally 8 columns), the longitudinal axis is that gradient direction corresponds to the cumulative of gradient magnitude,
In the principal direction that the peak value of histogram is exactly characteristic point.It also refers to carry out histogram using Gaussian function in the paper of Lowe
The smoothly effect with the close neighborhood point of Enhanced feature point to key point direction, and reduce the influence of mutation.It is more accurate in order to obtain
Direction, usually can also to discrete histogram of gradients carry out interpolation fitting.Specifically, the direction of key point can by with
Three nearest column values of main peak value are obtained by parabola interpolation.In histogram of gradients, main peak value is equivalent to when there are one
When the column value of 80% energy, then this direction being considered to, this feature point assists direction.
15% key point has multi-direction, and these points are very crucial to matched stability.
After obtaining the principal direction of characteristic point, available for each characteristic point three information (x, y, σ, θ) (x, y, σ,
θ), i.e. position, scale and direction.It is possible thereby to determine a SIFT feature region, a SIFT feature region is by three value tables
Show, center indicates that characteristic point position, radius indicate that the scale of key point, arrow indicate principal direction.Key with multiple directions
Point can be duplicated into more parts, then direction value is assigned to the characteristic point after duplication respectively, a characteristic point just produces multiple
Coordinate, scale are equal, but the characteristic point that direction is different.
Key point is described using one group of vector and namely generates feature point description, this descriptor not only includes feature
Point also contains around characteristic point to its contributive pixel.Description answers independence with higher, to guarantee matching rate.
Substantially there are three steps for the generation of feature descriptor:
(1) correction rotation principal direction, it is ensured that rotational invariance.
(2) description is generated, the feature vector of one 128 dimension is ultimately formed
(3) feature vector length is normalized normalized, further removes the influence of illumination.
In order to guarantee the rotational invariance of characteristic vector, nearby reference axis to be revolved in neighborhood centered on characteristic point
Turn θ (principal direction of characteristic point) angle, i.e., reference axis is rotated to be to the principal direction of characteristic point.After rotation in neighborhood pixel new seat
It is designated as: [x ' y ']=[cos θ-sin θ sin θ cos θ] [xy].
8 × 8 window is taken after rotation centered on principal direction.Center is the position of current key point, and each small lattice represent
For a pixel of scale space where crucial vertex neighborhood, the gradient magnitude and gradient direction of each pixel, arrow direction are sought
The gradient direction of the pixel is represented, then length representative gradient magnitude is weighted it using Gauss window.Finally exist
The histogram of gradients that 8 directions are drawn on each 4 × 4 fritter, calculates the accumulated value of each gradient direction, can form one
Seed point.Each characteristic point is made of 4 seed points, and each seed point has the vector information in 8 directions.This neighborhood directionality
Information consolidation enhances the noise resisting ability of algorithm, also provides simultaneously for the characteristic matching containing position error and compares rationality
Fault-tolerance.
From ask principal direction different, the histogram of gradients of each seed region is divided into 8 directions between 0-360 at this time
Section, each section are 45 degree, i.e., each seed point has the gradient intensity information in 8 directions.In actual calculating process, it is
Enhancing matched robustness,
To each key point using 4 × 4 totally 16 seed points describe, such a key point can generate 128 dimensions
SIFT feature vector.
By to around characteristic point pixel carry out piecemeal, calculation block inside gradient histogram, generate it is unique to
Amount, this vector are that one kind of the regional image information is abstract, have uniqueness.
Stereo matching refers to according to extracted feature a kind of corresponding relationship established between image pair, that is, will be same
Imaging point of one physical space o'clock in two width different images correspond.When being matched it is noted that in scene
The interference of some factors, such as the distortion of illumination condition, noise jamming, scene geometry, surface physical characteristic and video camera
Many changing factors such as machine characteristic.
Many Stereo Matching Algorithms are provided in opencv, similar to BM, global SGBM etc. of part, these calculations
Method is more probably that the faster effect of speed is poorer, if not investigating very much timeliness, and corrects that do is not good
Words are recommended to use SGBM.
First define a concept --- epipolar-line constraint, a point on left video camera, the point of corresponding three-dimensional spatially,
When we will look for this to put subpoint on the right, it is not necessary that this image is all traversed one time.
Step1:SGBM uses horizontal Sobel operator, image is processed, formula are as follows:
Sobel (x, y)=2 [P (x+1, y)-P (x-1, y)]+P (x+1, y-1)-P (x-1, y-1)+P (x+1, y+1)-P
(x-1,y+1)；
Step2: with a function will (P indicates it by each pixel on horizontal Sobel operator treated image
Pixel value) it is mapped to a new image: PNEW indicates the pixel value on new images.
Mapping function:
PreFilterCap is a constant parameter, takes 15 under opencv default condition, takes 63 in routine.
Pretreatment is actually to obtain the gradient information of image, pretreated image is saved, it will for calculating
Cost.
Cost is made of two parts:
1, the gradient information of the image obtained by pretreatment passes through the gradient cost that the method based on sampling obtains.
2, original image passes through the SAD cost that the method based on sampling obtains.
Plan formula:
Lr(p, d)=C (p, d)+min (Lr(p-r,d),
Lr(p-r,d-1)+P1,Lr(p-r,d+1)+P1,
Default 4 paths, wherein the critically important two parameters P1 of Dynamic Programming, P2 are set in this way:
P1=8*cn*sgbm.SADWindowSize*sgbm.SADWindowSize；
P2=32*cn*sgbm.SADWindowSize*sgbm.SADWindowSize；
Cn is the port number of image, and SADWindowSize is SAD window size, and numerical value is odd number.
As can be seen that the projecting parameter P1 and P2 of SGBM is constant when image channel and SAD window are decided.
The post-processing of opencvSGBM includes following steps:
Step1: uniqueness detection: lowest costs are time low-cost (1+ in parallax window ranges
UniquenessRatio/100) times when, the corresponding parallax value of lowest costs is only the parallax of the pixel, otherwise the pixel
Parallax be 0.Wherein uniquenessRatio is a constant parameter.
Step2: sub-pixel interpolation:
Step3: left and right consistency detection: error threshold disp12MaxDiff is defaulted as 1, can oneself setting.
OpencvSGBM calculates the mode of right disparity map: examining figure by obtained left view and calculates right disparity map
The thinking of SGBM: the disparity by choosing each pixel forms a disparity map
(disparitymap), a global energy function relevant with disparity map is set, minimizes this energy function, to reach
Solve the purpose of the optimal disparity of each pixel.
There is more accurate matching result, in conjunction with the inside and outside parameter of camera calibration, so that it may recover three-dimensional scenic
Information.Since reconstruction accuracy is by matching precision, the influence of the factors such as inside and outside parameter error of video camera, therefore firstly the need of
Carry out the work of the several steps in front so that the precision of links is high, error is small, can just design in this way one it is more accurate
Stereo visual system.The point cloud data of reconstruct output can be post-processed later, such as denoising, resurfacing.
Three-dimensional point cloud is post-processed, such as interpolation, trigonometric ratio, patch texture, finally obtains threedimensional model.
Second position determination unit, for obtaining human eye relative to using the cascade classifier integrated in OpenCV
Practise the position at end.The specific implementation principle of the unit are as follows:
Video is read from camera, carries out Face datection, further according to the facial image interception left eye and right eye detected
ROI region, it is last that eyeball Spot detection and tracking are carried out according to the ROI being truncated to.So algorithm mainly includes three parts:
Face datection, ROI interception, eyeball centralized positioning.
Here method for detecting human face is exactly the cascade classifier integrated in OpenCV, this is a kind of very ancient
Method for detecting human face form final classifier using Haar feature or the multiple weak typing models of LBP feature training, carry out
Face datection.
Multiple trained models are provided in OpenCV, it is contemplated that right and left eyes will hide in the case where side face
Gear, so only having used frontal faces detection at present.
The third place determination unit calculates human eye and study for the relative positional relationship according to study end and human eye
Hold the distance and angle of screen.The realization principle of the unit are as follows: human eye and study end screen are put into world coordinate system, according to
The two coordinate calculates distance and angle.
Accurate adjustment unit, for adjusting according to the calculated result of the third place unit and the corresponding relationship of the second setup unit
Screen intensity.
Technical characteristic of the present invention without description can realize that details are not described herein by or using the prior art, certainly,
The above description is not a limitation of the present invention, and the present invention is also not limited to the example above, the ordinary skill of the art
The variations, modifications, additions or substitutions that personnel are made within the essential scope of the present invention also should belong to protection model of the invention
It encloses.
Claims (9)
1. a kind of intelligent tutoring system deeply interacted, which is characterized in that the system includes:
Teaching end for sending interactive teaching and learning information to study end, and is looked into the authentication information and return information at study end
It askes；
Learn end, for realizing the biometric information authentication of student, and receives and reply the interactive information that teaching end is sent；
Storage end, for storing teaching end and learning the data information of end generation.
2. a kind of intelligent tutoring system deeply interacted according to claim 1, which is characterized in that the teaching end packet
It includes:
Information issuing module for uploading interactive teaching and learning information, and sends this information to each study end；
Enquiry module, authentication information and return information for query learning end.
3. a kind of intelligent tutoring system deeply interacted according to claim 1 or 2, which is characterized in that the study
End includes:
Authentication module carries out typing and identification certification for the biological information to each student；
Interactive module, for receiving the touch screen of the teaching interaction information and simultaneous display at teaching end at study end, interaction letter
Breath realizes presentation process according to the guide of demonstration instruction by the way of human assistance demonstration, by student.
4. a kind of intelligent tutoring system deeply interacted according to claim 3, which is characterized in that the study end is also
Include:
Course recommending module, by calculating the similarity of all previous student's electives, or the weight of each elective of analysis,
Student is assisted to carry out the selection of elective course.
5. a kind of intelligent tutoring system deeply interacted according to claim 4, which is characterized in that the course pushes away
Recommending module includes:
Similarity analysis unit, the unit are based on multiple view hash algorithm, encode to the historical data of all previous students
After obtain baseline encoded, real-time coding is then calculated according to the real time data of curricula-variable student, by comparing real-time coding and base
Quasi- coding, finds and the highest historical data of real time data similarity；Or
Weight analysis unit is based on comprehensive weight analytic approach, analyzes weight coefficient of each elective compared to specialized course,
It was found that influencing maximum elective course to specialized course.
6. a kind of intelligent tutoring system deeply interacted according to claim 3, which is characterized in that the study end is also
Include:
Brightness coarse adjustment module adjusts screen intensity for the corresponding relationship of illuminance and screen intensity by prestoring in real time.
7. a kind of intelligent tutoring system deeply interacted according to claim 6, which is characterized in that the brightness coarse adjustment
Module includes:
First setup unit, for setting the corresponding relationship of illuminance and screen intensity；
Illuminance measurement unit, for detecting the illuminance in the study end external world；
Adjust unit, the illuminance for obtaining according to illuminance measurement unit, the corresponding relationship tune set according to setup unit
Save screen intensity.
8. a kind of intelligent tutoring system deeply interacted according to claim 3, which is characterized in that the study end is also
Include:
Brightness accurate adjustment module, for adjusting screen intensity according to the relative positional relationship of human eye and screen.
9. a kind of intelligent tutoring system deeply interacted according to claim 8, which is characterized in that the brightness accurate adjustment
Module includes:
Second setup unit, for setting the corresponding relationship between human eye distance, angle and screen intensity three；
First position determination unit, for obtaining face relative to study end using binocular restructing algorithm on the basis of learning end
Position and posture；
Second position determination unit, for obtaining human eye relative to study end using the cascade classifier integrated in OpenCV
Position；
The third place determination unit calculates human eye and study end screen for the relative positional relationship according to study end and human eye
The distance and angle of curtain；
Accurate adjustment unit, for adjusting screen according to the calculated result of the third place unit and the corresponding relationship of the second setup unit
Brightness.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910655383.9A CN110533971A (en) | 2019-07-19 | 2019-07-19 | A kind of intelligent tutoring system deeply interacted |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910655383.9A CN110533971A (en) | 2019-07-19 | 2019-07-19 | A kind of intelligent tutoring system deeply interacted |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110533971A true CN110533971A (en) | 2019-12-03 |
Family
ID=68661860
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910655383.9A Pending CN110533971A (en) | 2019-07-19 | 2019-07-19 | A kind of intelligent tutoring system deeply interacted |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110533971A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112735198A (en) * | 2020-12-31 | 2021-04-30 | 深兰科技（上海）有限公司 | Experiment teaching system and method |
| CN113221798A (en) * | 2021-05-24 | 2021-08-06 | 南京伯索网络科技有限公司 | Classroom student aggressiveness evaluation system based on network |
| CN117152688A (en) * | 2023-10-31 | 2023-12-01 | 江西拓世智能科技股份有限公司 | Intelligent classroom behavior analysis method and system based on artificial intelligence |
Citations (27)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2004029905A1 (en) * | 2002-09-27 | 2004-04-08 | Ginganet Corporation | Remote education system, course attendance check method, and course attendance check program |
| WO2013049907A1 (en) * | 2010-10-07 | 2013-04-11 | Clevru Corporation | Method, system and computer program for providing an intelligent collaborative content infrastructure |
| CN103544663A (en) * | 2013-06-28 | 2014-01-29 | Tcl集团股份有限公司 | Method and system for recommending network public classes and mobile terminal |
| CN104008515A (en) * | 2014-06-04 | 2014-08-27 | 江苏金智教育信息技术有限公司 | Intelligent course selection recommendation method |
| CN104680453A (en) * | 2015-02-28 | 2015-06-03 | 北京大学 | Course recommendation method and system based on students' attributes |
| CN104735241A (en) * | 2015-02-12 | 2015-06-24 | 广东欧珀移动通信有限公司 | Mobile terminal and control method for screen brightness of mobile terminal |
| CN105138624A (en) * | 2015-08-14 | 2015-12-09 | 北京矩道优达网络科技有限公司 | Personalized recommendation method based on user data of on-line courses |
| CN106210536A (en) * | 2016-08-04 | 2016-12-07 | 深圳众思科技有限公司 | A kind of screen luminance adjustment method, device and terminal |
| CN106195247A (en) * | 2016-10-08 | 2016-12-07 | 深圳万发创新进出口贸易有限公司 | A kind of control system of speed variator based on big Db Management Model |
| CN106339829A (en) * | 2016-11-10 | 2017-01-18 | 国网山东省电力公司济南供电公司 | Big data, Cloud, IoT and mobile internet technologies based active maintenance panorama monitoring system of power distribution network |
| CN106485969A (en) * | 2016-12-20 | 2017-03-08 | 成都远策数码科技有限公司 | A kind of on-line teaching system and method |
| CN106503864A (en) * | 2016-11-10 | 2017-03-15 | 国网山东省电力公司济南供电公司 | A kind of classification prediction for supporting Distribution Network Failure actively to rush to repair and method for early warning |
| CN106779096A (en) * | 2016-11-10 | 2017-05-31 | 国网山东省电力公司济南供电公司 | Power distribution network reports situation active forewarning system for repairment |
| CN107958351A (en) * | 2017-12-26 | 2018-04-24 | 重庆大争科技有限公司 | Teaching quality assessment cloud service platform |
| CN108154451A (en) * | 2017-11-30 | 2018-06-12 | 陕西铁路工程职业技术学院 | A kind of intelligent teaching system |
| CN108876014A (en) * | 2018-05-29 | 2018-11-23 | 黑龙江省经济管理干部学院 | A kind of OBE education and instruction course optimization system |
| CN108965728A (en) * | 2018-07-06 | 2018-12-07 | 维沃移动通信有限公司 | A kind of brightness adjusting method and terminal |
| CN109165824A (en) * | 2018-08-07 | 2019-01-08 | 国网江苏省电力有限公司 | A kind of appraisal procedure and system for critical workflow |
| CN109191189A (en) * | 2018-08-20 | 2019-01-11 | 国网河南省电力公司经济技术研究院 | Power sales decontrol lower power customer value assessment method |
| CN109190897A (en) * | 2018-08-01 | 2019-01-11 | 安徽宇烁光电科技有限公司 | A kind of campus wisdom class board generalized information management system |
| CN109584662A (en) * | 2018-12-29 | 2019-04-05 | 沈阳体育学院 | A kind of ideological politics in university theory teaching system |
| CN109582864A (en) * | 2018-11-19 | 2019-04-05 | 华南师范大学 | Course recommended method and system based on big data science and changeable weight adjustment |
| CN109740863A (en) * | 2018-12-13 | 2019-05-10 | 国网山东省电力公司经济技术研究院 | Integrated evaluating method based on big plant-grid connection system |
| CN109754348A (en) * | 2019-01-03 | 2019-05-14 | 寂通(上海)管理咨询有限公司 | A kind of course method for pushing, device, electronic equipment and medium |
| CN109841103A (en) * | 2017-11-27 | 2019-06-04 | 天津祎智教育科技有限公司 | One kind is based on teacher, student, the comprehensive mathematical studying system of parent tripartite |
| CN109859550A (en) * | 2018-12-24 | 2019-06-07 | 凡悦科技(上海)有限公司 | A kind of education cloud platform |
| KR101993574B1 (en) * | 2019-01-14 | 2019-06-26 | 이은비 | A System Providing Multi-Lingual Diary Editing Service |
-
2019
- 2019-07-19 CN CN201910655383.9A patent/CN110533971A/en active Pending
Patent Citations (27)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2004029905A1 (en) * | 2002-09-27 | 2004-04-08 | Ginganet Corporation | Remote education system, course attendance check method, and course attendance check program |
| WO2013049907A1 (en) * | 2010-10-07 | 2013-04-11 | Clevru Corporation | Method, system and computer program for providing an intelligent collaborative content infrastructure |
| CN103544663A (en) * | 2013-06-28 | 2014-01-29 | Tcl集团股份有限公司 | Method and system for recommending network public classes and mobile terminal |
| CN104008515A (en) * | 2014-06-04 | 2014-08-27 | 江苏金智教育信息技术有限公司 | Intelligent course selection recommendation method |
| CN104735241A (en) * | 2015-02-12 | 2015-06-24 | 广东欧珀移动通信有限公司 | Mobile terminal and control method for screen brightness of mobile terminal |
| CN104680453A (en) * | 2015-02-28 | 2015-06-03 | 北京大学 | Course recommendation method and system based on students' attributes |
| CN105138624A (en) * | 2015-08-14 | 2015-12-09 | 北京矩道优达网络科技有限公司 | Personalized recommendation method based on user data of on-line courses |
| CN106210536A (en) * | 2016-08-04 | 2016-12-07 | 深圳众思科技有限公司 | A kind of screen luminance adjustment method, device and terminal |
| CN106195247A (en) * | 2016-10-08 | 2016-12-07 | 深圳万发创新进出口贸易有限公司 | A kind of control system of speed variator based on big Db Management Model |
| CN106779096A (en) * | 2016-11-10 | 2017-05-31 | 国网山东省电力公司济南供电公司 | Power distribution network reports situation active forewarning system for repairment |
| CN106503864A (en) * | 2016-11-10 | 2017-03-15 | 国网山东省电力公司济南供电公司 | A kind of classification prediction for supporting Distribution Network Failure actively to rush to repair and method for early warning |
| CN106339829A (en) * | 2016-11-10 | 2017-01-18 | 国网山东省电力公司济南供电公司 | Big data, Cloud, IoT and mobile internet technologies based active maintenance panorama monitoring system of power distribution network |
| CN106485969A (en) * | 2016-12-20 | 2017-03-08 | 成都远策数码科技有限公司 | A kind of on-line teaching system and method |
| CN109841103A (en) * | 2017-11-27 | 2019-06-04 | 天津祎智教育科技有限公司 | One kind is based on teacher, student, the comprehensive mathematical studying system of parent tripartite |
| CN108154451A (en) * | 2017-11-30 | 2018-06-12 | 陕西铁路工程职业技术学院 | A kind of intelligent teaching system |
| CN107958351A (en) * | 2017-12-26 | 2018-04-24 | 重庆大争科技有限公司 | Teaching quality assessment cloud service platform |
| CN108876014A (en) * | 2018-05-29 | 2018-11-23 | 黑龙江省经济管理干部学院 | A kind of OBE education and instruction course optimization system |
| CN108965728A (en) * | 2018-07-06 | 2018-12-07 | 维沃移动通信有限公司 | A kind of brightness adjusting method and terminal |
| CN109190897A (en) * | 2018-08-01 | 2019-01-11 | 安徽宇烁光电科技有限公司 | A kind of campus wisdom class board generalized information management system |
| CN109165824A (en) * | 2018-08-07 | 2019-01-08 | 国网江苏省电力有限公司 | A kind of appraisal procedure and system for critical workflow |
| CN109191189A (en) * | 2018-08-20 | 2019-01-11 | 国网河南省电力公司经济技术研究院 | Power sales decontrol lower power customer value assessment method |
| CN109582864A (en) * | 2018-11-19 | 2019-04-05 | 华南师范大学 | Course recommended method and system based on big data science and changeable weight adjustment |
| CN109740863A (en) * | 2018-12-13 | 2019-05-10 | 国网山东省电力公司经济技术研究院 | Integrated evaluating method based on big plant-grid connection system |
| CN109859550A (en) * | 2018-12-24 | 2019-06-07 | 凡悦科技(上海)有限公司 | A kind of education cloud platform |
| CN109584662A (en) * | 2018-12-29 | 2019-04-05 | 沈阳体育学院 | A kind of ideological politics in university theory teaching system |
| CN109754348A (en) * | 2019-01-03 | 2019-05-14 | 寂通(上海)管理咨询有限公司 | A kind of course method for pushing, device, electronic equipment and medium |
| KR101993574B1 (en) * | 2019-01-14 | 2019-06-26 | 이은비 | A System Providing Multi-Lingual Diary Editing Service |
Non-Patent Citations (1)
| Title |
|---|
| 曹秀丽: "协同过滤技术在高效选课推荐***中的应用研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112735198A (en) * | 2020-12-31 | 2021-04-30 | 深兰科技（上海）有限公司 | Experiment teaching system and method |
| CN113221798A (en) * | 2021-05-24 | 2021-08-06 | 南京伯索网络科技有限公司 | Classroom student aggressiveness evaluation system based on network |
| CN117152688A (en) * | 2023-10-31 | 2023-12-01 | 江西拓世智能科技股份有限公司 | Intelligent classroom behavior analysis method and system based on artificial intelligence |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109815893B (en) | Color face image illumination domain normalization method based on cyclic generation countermeasure network | |
| CN105956582B (en) | A kind of face identification system based on three-dimensional data | |
| CN107679537B (en) | A kind of texture-free spatial target posture algorithm for estimating based on profile point ORB characteristic matching | |
| CN103839277B (en) | A kind of mobile augmented reality register method of outdoor largescale natural scene | |
| CN104318569B (en) | Space salient region extraction method based on depth variation model | |
| CN110533971A (en) | A kind of intelligent tutoring system deeply interacted | |
| CN110059741A (en) | Image-recognizing method based on semantic capsule converged network | |
| CN105913487A (en) | Human eye image iris contour analyzing and matching-based viewing direction calculating method | |
| CN109118528A (en) | Singular value decomposition image matching algorithm based on area dividing | |
| CN110232389A (en) | A kind of stereoscopic vision air navigation aid based on green crop feature extraction invariance | |
| CN107563328A (en) | A kind of face identification method and system based under complex environment | |
| CN109584290A (en) | A kind of three-dimensional image matching method based on convolutional neural networks | |
| TW201042574A (en) | Image preprocessing system for 3D image database construction | |
| CN101398886A (en) | Rapid three-dimensional face identification method based on bi-eye passiveness stereo vision | |
| CN106780546B (en) | The personal identification method of motion blur encoded point based on convolutional neural networks | |
| CN109543632A (en) | A kind of deep layer network pedestrian detection method based on the guidance of shallow-layer Fusion Features | |
| CN106595702B (en) | A kind of multisensor spatial registration method based on astronomy calibration | |
| CN102982334B (en) | The sparse disparities acquisition methods of based target edge feature and grey similarity | |
| CN107292299B (en) | Side face recognition methods based on kernel specification correlation analysis | |
| CN106355195A (en) | The system and method used to measure image resolution value | |
| CN109087245A (en) | Unmanned aerial vehicle remote sensing image mosaic system based on neighbouring relations model | |
| CN110163567A (en) | Classroom roll calling system based on multitask concatenated convolutional neural network | |
| CN109035345A (en) | The TOF camera range correction method returned based on Gaussian process | |
| CN110532865A (en) | Spacecraft structure recognition methods based on visible light and laser fusion | |
| CN107704840A (en) | A kind of remote sensing images Approach for road detection based on deep learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191203 |