CN105631410B - A kind of classroom detection method based on intelligent video processing technique - Google Patents
A kind of classroom detection method based on intelligent video processing technique Download PDFInfo
- Publication number
- CN105631410B CN105631410B CN201510970928.7A CN201510970928A CN105631410B CN 105631410 B CN105631410 B CN 105631410B CN 201510970928 A CN201510970928 A CN 201510970928A CN 105631410 B CN105631410 B CN 105631410B
- Authority
- CN
- China
- Prior art keywords
- image
- classroom
- black
- detected
- frame
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/20—Education
- G06Q50/205—Education administration or guidance
Abstract
The present invention provides a kind of classroom detection methods based on intelligent video processing technique, which is characterized in that including demographic method of turning out for work；Demographic method of turning out for work include: extracted from classroom monitor video several material image-normalizeds-generation profiler-use Adaboost algorithm with student's upper half as identification target generate classifier-use scan child window traverse image to be detected-statistics always turn out for work number the step of.The classroom detection method can save classroom monitoring cost, and accurate statistics go out number of turning out for work in classroom, and statistical accuracy is high, realizes automated teaching quality evaluation, saves the time of calling the roll, and save human cost.
Description
Technical field
The present invention relates to classroom detection technique field, more specifically to a kind of based on intelligent video processing technique
Classroom detection method.
Background technique
Intelligent Video Surveillance Technology is derived from computer technology, digital image processing techniques and artificial intelligence technology, its benefit
With computer vision (Computer Vision, CV) and the method for video analysis (Video Analysis, VA) to video sequence
A series of analyses of row, realize detection, positioning, identification and the tracking to target in dynamic scene, and analyze and sentence on this basis
The behavior of disconnected target, so that daily management mission can be completed but also made a response in time when abnormal conditions occur.
Application based on existing intelligent video analysis is mainly gathered in abnormality detection, the people flow rate statistical etc. of video monitoring.
Wherein moving object detection, segmentation, recognition and tracking are Railway Projects relatively common in intelligent video analysis research field,
It is then a research emphasis problem very popular since the last few years as behavior understanding and descriptive analysis.
In school, the classrooms such as student attendance number, classroom discipline situation is all the importance of school control, is teaching
The important indicator of quality evaluation.Therefore teacher needs to call the roll on classroom to determine attendance, and school can also arrange supervisor
Make an inspection tour classroom；Consume many human resources.At this stage, camera is generally installed in school classroom and records classroom monitoring view
Frequently, but the effect of classroom monitor video is primarily to carry out security monitoring to classroom situation；Classroom monitor video is applied to
Classroom detection field is still in blank.At this stage, recognition of face is carried out frequently with the strong classifier that Haar+Adaboost is generated.
But the strong classifier that Haar+Adaboost is generated is for the face check frequency with the presence of certain tilt angle, detection effect
It is bad.And the shooting angle of camera is fixed in classroom, usual camera is supervised positioned at the oblique upper of student, therefore from classroom
Control video extraction to test object be the face for having certain tilt angle.It is found through experiments that, using Haar+Adaboost
The strong classifier of generation is unable to reach the requirement that number is detected from the monitor video of classroom.
Summary of the invention
It is an object of the invention to overcome shortcoming and deficiency in the prior art, provide a kind of based on intelligent video processing skill
Art, can save classroom monitoring cost, can accurate statistics go out in classroom that number of turning out for work, statistical accuracy are high, automation religion can be achieved
The classroom detection method for learning quality evaluation, saving the time of calling the roll, saving human cost.
In order to achieve the above object, the technical scheme is that: one kind based on intelligent video handle
The classroom detection method of technology, which is characterized in that including demographic method of turning out for work；Demographic method of turning out for work includes following step
It is rapid:
S1 step, obtains classroom monitor video, and several record images are extracted from classroom monitor video from database；From
The region for intercepting student's upper part of the body in image is recorded as material image one, and intercepts and does not include the region work for having student's upper part of the body
For material image two；Material image one and material image two are normalized respectively, to realize all material images one
It is identical with the equal size of material image two；
S2 step, if each material image one and material image two are divided into stem cell units respectively；It calculates separately each thin
The histogram of each pixel in born of the same parents' unit；Histogram is combined to the profiler and material for being respectively formed material image one
The profiler of image two；
S3 step, using the profiler of material image one as positive sample, using the profiler of material image two as
Negative sample；Positive sample and negative sample are learnt using Adaboost algorithm, generate classifier, the identification target of classifier is
Student is above the waist；
S4 step, obtains the corresponding classroom monitor video in classroom to be detected, from database to obtain image to be detected；It will be to
Detection image is pre-processed；Setting scanning child window makes image to be detected size and scans the ratio between child window size
For the initial value of setting ratio；Number value is set as zero；
S5 step traverses image to be detected using scanning child window and obtains several scanning subgraphs；Successively sentenced using classifier
Each scanning subgraph that breaks whether be classifier identification target: if the identification target of classifier, then number value from plus one；It is no
Then number value is constant；
S6 step, judge whether to have traversed all setting ratios: if having traversed all setting ratios, current number value is
It always turns out for work number, demographics of turning out for work EP (end of program)；Otherwise image to be detected size and/or scanning child window size are adjusted, is made
Ratio between image to be detected size and scanning child window size is next setting ratio, and skips to S5 step.
Classroom detection method of the present invention, classroom monitor video school is generally existing, that be used for security monitoring carry out class
Hall analysis, is the development and utilization to existing resource, can save classroom monitoring cost.Classroom detection method of the present invention can accurate statistics
It turns out for work in classroom out number, realizes automated teaching quality evaluation, save the time of calling the roll, save human cost.It is supervised using classroom
The fixed feature of video photography angle is controlled, the learning sample using the material image of classroom monitor video extraction as classifier,
Compared with existing classifier, the classifier that classroom detection method of the present invention trains can more effectively, quickly and accurately realize knowledge
Not.The identification goal-setting of classifier is student's upper part of the body, and student refers to the position of student's shoulder or more above the waist；This is because
Classroom middle school student block lower half portion by desk, are exposed to the part under camera using student as much as possible in the present invention；
Compared with only identifying face, recognition accuracy is can be improved in identification student above the waist.
Further embodiment is: classroom detection method further includes classroom discipline detection method；The classroom discipline detection side
Method includes the following steps:
T1 step, obtain classroom at the beginning of and the end time, obtain from the outset between to the end time classroom monitor
Each frame image in video；The spy of each pixel in image corresponding to the time started is characterized using gauss hybrid models
Sign；Next frame image is set as present analysis image；
T2 step, each pixel in present analysis image is matched with gauss hybrid models: if successful match,
Determine the pixel for background dot；Otherwise determine the pixel for foreground point；Using present analysis image update Gaussian Mixture mould
Type；
T3 step, judges whether the present analysis image corresponding time is the end time: if so, skipping to T4 step；Otherwise it sets
Next frame image is determined as present analysis image, and skips to T2 step；
T4 step, deletes background dot respectively in each frame image to form each frame foreground image；Respectively by each frame foreground image
Binaryzation obtains each frame black and white foreground image；Using function cvFindContours in OpenCV to each frame black and white foreground image into
Row processing, obtains the moving target quantity in each frame black and white foreground image；
T5 step, judge respectively moving target quantity in each frame black and white foreground image and moving target quantity setting limit value it
Between size: if moving target quantity in the frame black and white foreground image >=moving target quantity sets limit value, determine the frame
Black and white foreground image is abnormal black and white foreground image；Otherwise determine that the frame black and white foreground image is positive normally-black white foreground image；
T6 step counts the frequency of occurrences of abnormal black and white foreground image, and continuously several frame black and white foreground images are equal for statistics
For the Abnormal lasting of abnormal black and white foreground image, longest Abnormal lasting is found out；Judge abnormal black and white foreground image
The frequency of occurrences and longest Abnormal lasting: if the frequency of occurrences of abnormal black and white foreground image >=frequency setting limit value or longest
Abnormal lasting >=duration sets limit value, then judges the classroom discipline for abnormality；Otherwise determine the classroom discipline
For normal condition.
The also detectable classroom discipline situation of classroom detection method of the present invention, is analyzed using the image of classroom monitor video
Judgement, testing cost is low, can save classroom monitoring cost, and the monitor mode in classroom is maked an inspection tour without supervisor, saves manpower
Cost.In classroom detection method of the present invention, judged in conjunction with specific behaviors feature of walking about etc. on a large scale, using letter in OpenCV
Number cvFindContours handle each frame black and white foreground image, obtain moving target quantity and carry out subsequent judgement, can
Improve judging nicety rate.
Preferable scheme is that: in the S2 step, histogram refers to gradient orientation histogram or edge orientation histogram.
In the S4 step, image to be detected is subjected to pretreatment and is referred to, image to be detected is subjected to reduction noise, compensation light
According to processing.Due to classroom monitor video record quality it is irregular, carry out reduction noise, compensation lighting process can be to figure to be detected
As optimizing, image to be detected is made to can satisfy subsequent processing requirement.
In the S6 step, adjusts image to be detected size and/or scanning child window size refers to, using following three kinds of situations
One of: one, zoom in or out scanning child window size, image to be detected size constancy；Two, image to be detected size is reduced, is swept
Retouch child window size constancy；Three, magnified sweep child window size reduces image to be detected size.
In the T1 step, the spy of each pixel in image corresponding to the time started is characterized using gauss hybrid models
Sign refers to, the feature of each pixel in image corresponding to the time started is characterized using k Gauss model.
In the T4 step, carrying out processing to each frame black and white foreground image using function cvFindContours in OpenCV is
Refer to, moving target profile is searched using function cvFindContours in OpenCV, deletes area and be less than contour area setting value
Moving target profile, calculate residual movement objective contour quantity；Residual movement objective contour quantity is moving target quantity.This
Sample processing can avoid converting caused by environmental factor and causing to judge by accident.
Compared with prior art, the invention has the advantages that with the utility model has the advantages that
1, classroom detection method of the present invention is generally existing by school, is used for the classroom monitor video of security monitoring to carry out class
Hall analysis, is the development and utilization to existing resource, can save classroom monitoring cost；Classroom detection method of the present invention can accurate statistics
It turns out for work in classroom out number, realizes automated teaching quality evaluation, save the time of calling the roll, save human cost；
2, the feature that classroom detection method of the present invention utilizes monitor video camera angle in classroom fixed is monitored using classroom and is regarded
Learning sample of the material image that frequency extracts as classifier, compared with existing classifier, detection method training in classroom of the present invention
Classifier out can more effectively, quickly and accurately realize identification；The identification goal-setting of classifier is student's upper part of the body, student
Refer to the position of student's shoulder or more above the waist；This is because classroom middle school student block lower half portion by desk, use up in the present invention
The part under camera may be exposed to using student morely；Compared with only identifying face, identification student can be improved above the waist
Recognition accuracy；
3, classroom detection method of the present invention can detect classroom discipline situation, be analyzed using the image of classroom monitor video
Judgement, testing cost is low, can save classroom monitoring cost, and the monitor mode in classroom is maked an inspection tour without supervisor, saves manpower
Cost.
Detailed description of the invention
Fig. 1 is the flow chart of demographic method of turning out for work in classroom detection method of the present invention；
Fig. 2 is the flow chart of classroom discipline detection method in classroom detection method of the present invention.
Specific embodiment
The present invention is described in further detail with specific embodiment with reference to the accompanying drawing.
Embodiment
Classroom detection method of the present embodiment based on intelligent video processing technique, including demographic method of turning out for work；It turns out for work
The process of demographic method is as shown in Figure 1, include the following steps:
S1 step, obtains classroom monitor video, and several record images are extracted from classroom monitor video from database；From
The region for intercepting student's upper part of the body in image is recorded as material image one, and intercepts and does not include the region work for having student's upper part of the body
For material image two；Material image one and material image two are normalized respectively, to realize all material images one
It is identical with the equal size of material image two；
S2 step, if each material image one and material image two are divided into stem cell units respectively；It calculates separately each thin
The histogram of each pixel in born of the same parents' unit；Histogram is combined to the profiler and material for being respectively formed material image one
The profiler of image two；
S3 step, using the profiler of material image one as positive sample, using the profiler of material image two as
Negative sample；Positive sample and negative sample are learnt using Adaboost algorithm, generate classifier, the identification target of classifier is
Student is above the waist；
S4 step, obtains the corresponding classroom monitor video in classroom to be detected, from database to obtain image to be detected；It will be to
Detection image is pre-processed；Setting scanning child window makes image to be detected size and scans the ratio between child window size
For the initial value of setting ratio；Number value is set as zero；
S5 step traverses image to be detected using scanning child window and obtains several scanning subgraphs；Successively sentenced using classifier
Each scanning subgraph that breaks whether be classifier identification target: if the identification target of classifier, then number value from plus one；It is no
Then number value is constant；
S6 step, judge whether to have traversed all setting ratios: if having traversed all setting ratios, current number value is
It always turns out for work number, demographics of turning out for work EP (end of program)；Otherwise image to be detected size and/or scanning child window size are adjusted, is made
Ratio between image to be detected size and scanning child window size is next setting ratio, and skips to S5 step.
Classroom detection method of the present invention, classroom monitor video school is generally existing, that be used for security monitoring carry out class
Hall analysis, is the development and utilization to existing resource, can save classroom monitoring cost.Classroom detection method of the present invention can accurate statistics
It turns out for work in classroom out number, counts class attendance rate, realize automated teaching quality evaluation, save the time of calling the roll, save manpower
Cost.Using the feature that classroom monitor video camera angle is fixed, it is used as and is divided using the material image that classroom monitor video extracts
The learning sample of class device, compared with existing classifier, the classifier that classroom detection method of the present invention trains can be more effective, fast
Speed is accurately realized identification.The identification goal-setting of classifier is student's upper part of the body, and student refers to student's shoulder or more above the waist
Position；This is because classroom middle school student block lower half portion by desk, it is exposed to as much as possible using student in the present invention
Part under camera；Compared with only identifying face, recognition accuracy is can be improved in identification student above the waist.
Wherein, in the S2 step, histogram refers to gradient orientation histogram or edge orientation histogram.S2 step essence be into
Row image feature selection and extraction；Since the data volume of image is sizable, so needing real using the method for feature extraction
The compression of existing data.The purpose of feature extraction is to convert initial data, obtains the feature that can most reflect target essence.
In follow-up study, data can be replaced with feature.In the present invention, realized using HOG feature, this is because from HOG feature
Video material in can collect a large amount of useful data in relation to fixed angle face, be conducive to subsequent classifier training.
In S3 step, Adaboost algorithm is a kind of iterative algorithm, and core concept is trained not for the same training set
Same classifier (Weak Classifier), then gets up these weak classifier sets, constitutes (strong point of a stronger final classification device
Class device).Its algorithm itself is realized by changing data distribution, it is according to the classification of each sample among each training set
Whether the accuracy rate of correct and last time general classification, to determine the weight of each sample.The new data of weight will be modified
Collection is given sub-classification device and is trained, and finally finally merges the classifier that each training obtains, and determines as last
Plan classifier.Some unnecessary training data features can be excluded using the classifier that Adaboost algorithm is formed, and are placed on
Above crucial training data.
In the S4 step, image to be detected is subjected to pretreatment and is referred to, image to be detected is subjected to reduction noise, compensation light
According to processing.Due to classroom monitor video record quality it is irregular, carry out reduction noise, compensation lighting process can be to figure to be detected
As optimizing, image to be detected is made to can satisfy subsequent processing requirement.
In the S6 step, adjusts image to be detected size and/or scanning child window size refers to, using following three kinds of situations
One of: one, zoom in or out scanning child window size, image to be detected size constancy；Two, image to be detected size is reduced, is swept
Retouch child window size constancy；Three, magnified sweep child window size reduces image to be detected size.
It in practical applications, can also be using the image in addition to classroom as negative sample.
Classroom detection method of the present invention further includes classroom discipline detection method；The process of the classroom discipline detection method is such as
Shown in Fig. 2, include the following steps:
T1 step, obtain classroom at the beginning of and the end time, obtain from the outset between to the end time classroom monitor
Each frame image in video；The spy of each pixel in image corresponding to the time started is characterized using gauss hybrid models
Sign；Next frame image is set as present analysis image；
T2 step, each pixel in present analysis image is matched with gauss hybrid models: if successful match,
Determine the pixel for background dot；Otherwise determine the pixel for foreground point；Using present analysis image update Gaussian Mixture mould
Type；
T3 step, judges whether the present analysis image corresponding time is the end time: if so, skipping to T4 step；Otherwise it sets
Next frame image is determined as present analysis image, and skips to T2 step；
T4 step, deletes background dot respectively in each frame image to form each frame foreground image；Respectively by each frame foreground image
Binaryzation obtains each frame black and white foreground image；Using function cvFindContours in OpenCV to each frame black and white foreground image into
Row processing, obtains the moving target quantity in each frame black and white foreground image；
T5 step, judge respectively moving target quantity in each frame black and white foreground image and moving target quantity setting limit value it
Between size: if moving target quantity in the frame black and white foreground image >=moving target quantity sets limit value, determine the frame
Black and white foreground image is abnormal black and white foreground image；Otherwise determine that the frame black and white foreground image is positive normally-black white foreground image；
T6 step counts the frequency of occurrences of abnormal black and white foreground image, and continuously several frame black and white foreground images are equal for statistics
For the Abnormal lasting of abnormal black and white foreground image, longest Abnormal lasting is found out；Judge abnormal black and white foreground image
The frequency of occurrences and longest Abnormal lasting: if the frequency of occurrences of abnormal black and white foreground image >=frequency setting limit value or longest
Abnormal lasting >=duration sets limit value, then judges the classroom discipline for abnormality；Otherwise determine the classroom discipline
For normal condition.
The also detectable classroom discipline situation of classroom detection method of the present invention, is analyzed using the image of classroom monitor video
Judgement, testing cost is low, can save classroom monitoring cost, and the monitor mode in classroom is maked an inspection tour without supervisor, saves manpower
Cost；Dull, the fixed feature based on classroom background utilizes stable background, effectively progress motion target tracking.Class of the present invention
In hall detection method, judged in conjunction with specific behaviors feature of walking about etc. on a large scale, using function in OpenCV
CvFindContours handles each frame black and white foreground image, obtains moving target quantity to carry out subsequent judgement, can mention
High judging nicety rate.
In the T1 step, the spy of each pixel in image corresponding to the time started is characterized using gauss hybrid models
Sign refers to, the feature of each pixel in image corresponding to the time started is characterized using k Gauss model.Gaussian Mixture mould
Type mainly has two parameters of variance and mean value to determine that different study mechanisms is taken in the study to mean value and variance, will be direct
Influence the stability, accuracy and convergence of gauss hybrid models.Due to being the background extracting modeling to moving target,
It needs to two parameter real-time updates of variance in gauss hybrid models and mean value.To improve under busy scene, big and slow movement
The detection effect of target can introduce the concept of weight mean value, establish gauss hybrid models and real-time update, then click through to pixel
The classification of row foreground point and background dot.
In the T4 step, carrying out processing to each frame black and white foreground image using function cvFindContours in OpenCV is
Refer to, moving target profile is searched using function cvFindContours in OpenCV, deletes area and be less than contour area setting value
Moving target profile, calculate residual movement objective contour quantity；Residual movement objective contour quantity is moving target quantity.It looks into
Look for moving target profile that can be to look for facial contour；Judge whether the area of moving target profile is greater than or equal to profile later
Area, is less than the movement of contour area setting value by area setting (such as contour area setting value is 40 pixel *, 40 pixel)
Objective contour is deleted；Processing can avoid converting caused by environmental factor and causing to judge by accident in this way.
The above embodiment is a preferred embodiment of the present invention, but embodiments of the present invention are not by above-described embodiment
Limitation, other any changes, modifications, substitutions, combinations, simplifications made without departing from the spirit and principles of the present invention,
It should be equivalent substitute mode, be included within the scope of the present invention.
Claims (6)
1. a kind of classroom detection method based on intelligent video processing technique, which is characterized in that including demographic method of turning out for work；
Demographic method of turning out for work includes the following steps:
S1 step, obtains classroom monitor video, and several record images are extracted from classroom monitor video from database；From record
The region of student's upper part of the body is intercepted in image as material image one, and is intercepted and be used as element not comprising the region for having student's upper part of the body
Material image two；Material image one and material image two are normalized respectively, to realize all material images one and element
The equal size of material image two is identical；
S2 step, if each material image one and material image two are divided into stem cell units respectively；Calculate separately each cell list
The histogram of each pixel in member；Histogram is combined to the profiler and material image for being respectively formed material image one
Two profiler；
S3 step, using the profiler of material image one as positive sample, using the profiler of material image two as negative sample
This；Positive sample and negative sample are learnt using Adaboost algorithm, generate classifier, the identification target of classifier is student
Above the waist；
S4 step, obtains the corresponding classroom monitor video in classroom to be detected, from database to obtain image to be detected；It will be to be detected
Image is pre-processed；Setting scanning child window, sets image to be detected size and the ratio scanned between child window size
The initial value of certainty ratio；Number value is set as zero；
S5 step traverses image to be detected using scanning child window and obtains several scanning subgraphs；Successively judged using classifier each
A scanning subgraph whether be classifier identification target: if the identification target of classifier, then number value from plus one；Otherwise people
Numerical value is constant；
S6 step, judges whether to have traversed all setting ratios: if having traversed all setting ratios, current number value is always to go out
Diligent number, demographics of turning out for work EP (end of program)；Otherwise image to be detected size and/or scanning child window size are adjusted, is made to be checked
Ratio between altimetric image size and scanning child window size is next setting ratio, and skips to S5 step；
It further include classroom discipline detection method；The classroom discipline detection method includes the following steps:
T1 step, obtain classroom at the beginning of and the end time, obtain from the outset between to the end time classroom monitor video
In each frame image；The feature of each pixel in image corresponding to the time started is characterized using gauss hybrid models；If
Next frame image is determined as present analysis image；
T2 step, each pixel in present analysis image is matched with gauss hybrid models: if successful match, being determined
The pixel is background dot；Otherwise determine the pixel for foreground point；Using present analysis image update gauss hybrid models；
T3 step, judges whether the present analysis image corresponding time is the end time: if so, skipping to T4 step；Otherwise under setting
One frame image skips to T2 step as present analysis image；
T4 step, deletes background dot respectively in each frame image to form each frame foreground image；Respectively by each frame foreground image two-value
Change and obtains each frame black and white foreground image；Using function cvFindContours in OpenCV to each frame black and white foreground image at
Reason, obtains the moving target quantity in each frame black and white foreground image；
T5 step is judged respectively between the moving target quantity in each frame black and white foreground image and moving target quantity setting limit value
Size: if moving target quantity >=moving target quantity in the frame black and white foreground image sets limit value, determine the frame black and white
Foreground image is abnormal black and white foreground image；Otherwise determine that the frame black and white foreground image is positive normally-black white foreground image；
T6 step, counts the frequency of occurrences of abnormal black and white foreground image, and it is different for counting continuous several frame black and white foreground images
The Abnormal lasting of normally-black white foreground image, finds out longest Abnormal lasting；Judge the appearance of abnormal black and white foreground image
Frequency and longest Abnormal lasting: if the frequency of occurrences of abnormal black and white foreground image >=frequency setting limit value or longest are abnormal
Duration >=duration sets limit value, then judges the classroom discipline for abnormality；Otherwise determine that the classroom discipline is positive
Normal state.
2. the classroom detection method according to claim 1 based on intelligent video processing technique, which is characterized in that the S2
In step, histogram refers to gradient orientation histogram or edge orientation histogram.
3. the classroom detection method according to claim 1 based on intelligent video processing technique, which is characterized in that the S4
In step, image to be detected is subjected to pretreatment and is referred to, image to be detected is subjected to reduction noise, compensation lighting process.
4. the classroom detection method according to claim 1 based on intelligent video processing technique, which is characterized in that the S6
In step, adjust image to be detected size and/or scanning child window size refer to, using one of following three kinds of situations: one, amplification or
Reduce scanning child window size, image to be detected size constancy；Two, image to be detected size is reduced, scanning child window size is not
Become；Three, magnified sweep child window size reduces image to be detected size.
5. the classroom detection method according to claim 1 based on intelligent video processing technique, which is characterized in that the T1
In step, the feature that each pixel in image corresponding to the time started is characterized using gauss hybrid models is referred to, using k
Gauss model characterizes the feature of each pixel in image corresponding to the time started.
6. the classroom detection method according to claim 1 based on intelligent video processing technique, which is characterized in that the T4
In step, processing is carried out to each frame black and white foreground image using function cvFindContours in OpenCV and is referred to, using OpenCV
Middle function cvFindContours searches moving target profile, deletes the moving target wheel that area is less than contour area setting value
Exterior feature calculates residual movement objective contour quantity；Residual movement objective contour quantity is moving target quantity.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510970928.7A CN105631410B (en) | 2015-12-18 | 2015-12-18 | A kind of classroom detection method based on intelligent video processing technique |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510970928.7A CN105631410B (en) | 2015-12-18 | 2015-12-18 | A kind of classroom detection method based on intelligent video processing technique |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105631410A CN105631410A (en) | 2016-06-01 |
| CN105631410B true CN105631410B (en) | 2019-04-09 |
Family
ID=56046323
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510970928.7A Active CN105631410B (en) | 2015-12-18 | 2015-12-18 | A kind of classroom detection method based on intelligent video processing technique |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105631410B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108109220A (en) * | 2017-12-29 | 2018-06-01 | 贵州理工学院 | A kind of classroom work attendance statistics system based on monitoring camera |
| CN108257056A (en) * | 2018-01-23 | 2018-07-06 | 余绍志 | A kind of classroom assisted teaching system for the big data for being applied to teaching industry |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2009301104A (en) * | 2008-06-10 | 2009-12-24 | Chube Univ | Object detecting apparatus |
| CN103699888A (en) * | 2013-12-29 | 2014-04-02 | 深圳市捷顺科技实业股份有限公司 | Human face detection method and device |
| CN104778453A (en) * | 2015-04-02 | 2015-07-15 | 杭州电子科技大学 | Night pedestrian detection method based on statistical features of infrared pedestrian brightness |
-
2015
- 2015-12-18 CN CN201510970928.7A patent/CN105631410B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2009301104A (en) * | 2008-06-10 | 2009-12-24 | Chube Univ | Object detecting apparatus |
| CN103699888A (en) * | 2013-12-29 | 2014-04-02 | 深圳市捷顺科技实业股份有限公司 | Human face detection method and device |
| CN104778453A (en) * | 2015-04-02 | 2015-07-15 | 杭州电子科技大学 | Night pedestrian detection method based on statistical features of infrared pedestrian brightness |
Non-Patent Citations (1)
| Title |
|---|
| 基于Adaboost算法和背景减法的电梯减法的电梯轿厢内人数检测研究;蔡俊;《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》;20140815;正文第28-33、40-41页 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105631410A (en) | 2016-06-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111444821B (en) | Automatic identification method for urban road signs | |
| CN106570486B (en) | Filtered target tracking is closed based on the nuclear phase of Fusion Features and Bayes's classification | |
| CN109948447B (en) | Character network relation discovery and evolution presentation method based on video image recognition | |
| Lim et al. | Automated classroom monitoring with connected visioning system | |
| CN102214291B (en) | Method for quickly and accurately detecting and tracking human face based on video sequence | |
| CN105469105A (en) | Cigarette smoke detection method based on video monitoring | |
| CN110728252B (en) | Face detection method applied to regional personnel motion trail monitoring | |
| CN109298785A (en) | A kind of man-machine joint control system and method for monitoring device | |
| CN112183238B (en) | Remote education attention detection method and system | |
| CN103034838A (en) | Special vehicle instrument type identification and calibration method based on image characteristics | |
| CN104123543A (en) | Eyeball movement identification method based on face identification | |
| CN108256462A (en) | A kind of demographic method in market monitor video | |
| CN107392251B (en) | Method for improving target detection network performance by using classified pictures | |
| CN103065163B (en) | A kind of fast target based on static images detects recognition system and method | |
| CN104123714A (en) | Optimal target detection scale generation method in people flow statistics | |
| CN114463843A (en) | Multi-feature fusion fish abnormal behavior detection method based on deep learning | |
| CN105631410B (en) | A kind of classroom detection method based on intelligent video processing technique | |
| CN110232306A (en) | A kind of present status system based on image detection | |
| CN109472228A (en) | A kind of yawn detection method based on deep learning | |
| Zeng et al. | Finger-vein image quality assessment based on light-cnn | |
| CN109766848B (en) | Pedestrian vision condition investigation and analysis method based on monitoring video | |
| CN111694980A (en) | Robust family child learning state visual supervision method and device | |
| Pei et al. | Convolutional neural networks for class attendance | |
| CN106650824B (en) | Moving object classification method based on support vector machines | |
| CN115424352A (en) | Method for identifying kitchen pest invasion based on computer vision |
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 |