CN106997587A - A kind of measuring method of the intravenous fluid drip speed based on machine vision - Google Patents
A kind of measuring method of the intravenous fluid drip speed based on machine vision Download PDFInfo
- Publication number
- CN106997587A CN106997587A CN201710178754.XA CN201710178754A CN106997587A CN 106997587 A CN106997587 A CN 106997587A CN 201710178754 A CN201710178754 A CN 201710178754A CN 106997587 A CN106997587 A CN 106997587A
- Authority
- CN
- China
- Prior art keywords
- frame
- drop
- pixel
- difference
- template
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
- G06T7/0014—Biomedical image inspection using an image reference approach
- G06T7/0016—Biomedical image inspection using an image reference approach involving temporal comparison
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01P—MEASURING LINEAR OR ANGULAR SPEED, ACCELERATION, DECELERATION, OR SHOCK; INDICATING PRESENCE, ABSENCE, OR DIRECTION, OF MOVEMENT
- G01P5/00—Measuring speed of fluids, e.g. of air stream; Measuring speed of bodies relative to fluids, e.g. of ship, of aircraft
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
Abstract
The present invention relates to a kind of measuring method of the intravenous fluid drip speed based on machine vision.Sample video is shot first, and video switchs to still image frame by frame;Then dropper is considered as moving target.Foreground target is extracted, and carries out binary conversion treatment;The frame number containing drop is extracted again, threshold value Th is set, when a certain frame pixel value is 255 and Sum > Th, then it is assumed that this frame is containing drop frame.Otherwise it is assumed that drop is not present in this frame.Flow velocity is calculated according to V=fps/ Δs N.Instant invention overcomes the not enough shortcoming that existing determination techniques are complicated, supervision is inconvenient, time-consuming, expensive.Intravenous fluid drip speed is easily measured for more precise and high efficiency, the workload for mitigating medical worker provides support;Shake of the present invention to video has carried out the processing of correlation, it is allowed to unstable in shooting process, it is not necessary to establishing shot equipment deliberately.Mobile terminal is transplanted to algorithm, and the accuracy of result brings breakthrough.
Description
Technical field
The invention belongs to image procossing and Intelligent Recognition field, and in particular to a kind of intravenous fluid based on machine vision
The measuring method of drip speed.
Background technology
So far, venous transfusion technology has developed last 600 years, but its it is real form complete set transfusion system be
In 20th century, turn into one of most effective, direct and conventional clinical medicine treatment means at present.Drip-feed is by woven hose
Substantial amounts of liquid and medicine are inputted into internal method by vein.For the medicine for being difficult to absorb, and vomiting, the disease of stupor
People, can be administered using the method for drip-feed.Its advantage is to absorb rapid, and dosage is accurate, reliable effect, and medicine is without stomach
Intestines and liver and be directly entered in tissue and body fluid, school assignment quick, be readily applicable to first aid and be unable to the patient of oral drugs.
In clinical medicine, transfusion is time-consuming more, may all carry out in daylight and at night, and situation particularly vein of infusing drips
Speed, it is necessary to observation in time, so as to according to drug type control drop speed, change medicine, transfusion at the end of pull out pin etc. in time, drop speed is surveyed
Amount medical personnel bring heavy burden, realize the automatic monitoring of transfusion and turn into the active demand of clinical practice.
Therefore, drip-feed is tested the speed for clinical treatment and medical research all significances.Traditional drip-feed is surveyed
The mode of flow velocity mainly has following several:(1) mechanical weighing formula infusion test;(2) infrared light electric-type infusion test;(3) electric capacity
Metering-type infusion test.
Traditional drip-feed, which is tested the speed, relies primarily on medical personnel's manual measurement, and manual measurement has many shortcomings:(1)
Detection speed is excessively slow, it is necessary to expend a large amount of manpowers;(2) testing result is inaccurate, and medical personnel often rely on personal experience or letter
Single timer detection drip-feed speed, as a result may be not accurate enough.Therefore, the method for Traditional Man detection is difficult in adapt to face
The development of bed medical research, necessity is become using the higher detection method of new, automaticity.
With the development of digital image processing techniques, it is also in an increasingly wide range of applications in medical domain, utilizes number
Word image processing method tests the speed to carry out drip-feed, can not only improve detection efficiency, increases the accuracy of testing result, together
When it is with low cost, automaticity is higher.
The content of the invention
The above-mentioned technical problem of the present invention is mainly what is be addressed by following technical proposals:
A kind of intravenous fluid drop flow-speed measurement method based on machine vision, it is characterised in that including:
Step 1, Sample video is shot, video switchs to still image frame by frame;
Step 2, dropper is considered as moving target.Foreground target is extracted, and carries out binary conversion treatment;
Step 3, extract the frame number containing drop, set threshold value Th, when a certain frame pixel value for 255 and Sum > Th, then
It is containing drop frame to think this frame.Otherwise it is assumed that drop is not present in this frame.Record the frame number containing drop.Different veins
Injection, due to the difference of switch controlled number, flow velocity is different.So there is the more than frame of frame of same dropping liquid drop or having
Situation about missing, records each frame and meets the numbering containing drop frame threshold value.Then these numberings are traveled through, it is continuous when there is numbering
Situation, then it is assumed that be same frame, calculates the average of serial number
Step 4, flow velocity is calculated according to V=fps/ Δs N.The average frame number of the appearance drop calculated first by above-mentioned stepsThe then adjacent difference of drop frameThe then average of frame-to-frame differencesFor Δ N1, Δ N2, Δ N3... average value, then flow velocityFps is the frame per second of video.
In the specific processing of step 2 described in a kind of above-mentioned intravenous fluid drop flow-speed measurement method based on machine vision
Step is based on two kinds of speed-measuring methods, including:
Speed-measuring method one, template matching method is specifically included:
Step 2.1.1, first have to extract drop template, the present invention uses Hough transformation, the method for circular extraction.Liquid
It is approximately circular to drop in form in drip irrigation.Extract after circle, it is (X1, Y1), radius of circle R to choose center of circle O coordinates.Then with coordinate
(X1-R, Y1-R) central point, length and width 2R interception drop templates.
Step 2.1.2, the drop of above onestep extraction are template, and template matches are carried out frame by frame to all frame of video.Typically
The algorithm idea of template:Search pattern T ((m × n) individual pixel) is overlayed and translated on searched figure S (W × H pixel), mould
That block region whistle figure S of the searched figure of plate coveringij.I, j are coordinate of the subgraph upper left corner on searched figure S.Hunting zone
It is:
1≤i≤W-m
1≤j≤W-m
By comparing T and SijSimilitude, complete template matches process.
Step 2.1.3, matching area binaryzation, because the frame containing drop and the difference of the frame without drop are than larger, so
Here binarization method is more random.Can be local binarization, global binaryzation can also be self-adaption binaryzation.Two-value
It is to preferably carry out the statistics of pixel to change purpose.
Step 2.1.4, the pixel Data-Statistics Sum of target area.Start from left to right from (0,0), from top to bottom progressively time
Image is gone through, the cumulative of pixel sum is carried out, Jia 1 if being 255 if traversal point pixel value, by that analogy.Calculate pixel value for 255 it is total
With.
Speed-measuring method two, frame difference method is specifically included:
Step 2.2.1, utilize GMM mixed Gauss models extract prospect.First by the average of each Gauss, variance, weights
0 is both configured to, that is, initializes a model matrix parameter.It is used for training GMM model using the T frames in video.To each pixel
For, set up the GMM model of the maximum GMM_MAX_COMPONT Gauss of its Number of Models.When first pixel is come, it is individually for
The initial mean value of its fixation, variance are set in a program, and weights are set to 1.
In non-first frame training process, when tail pixel value, with the average ratio of above existing Gauss compared with, if
The value of the pixel is with the equal value difference of its model in 3 times of variance, then task belongs to the Gauss.Now carried out more with equation below
Newly:
Whereinα=1/T,
Taken when the value of the pixel and the difference of average be not in the range of its 3 times
After the frame number T for reaching training, the selection of different pixels point GMM numbers adaptively is carried out.Weights divided by variance are used first
Each Gauss is sorted from big to small, the Gauss of foremost B is then chosen, meets it
Wherein CfGenerally 0.3
It thus can be very good to eliminate the noise spot in training process.In test phase, to the value and B of pixel of newly arriving
Each average in individual Gauss is compared, if between variance of its difference at 2 times, then it is assumed that be background, otherwise
It is considered prospect.As long as and wherein thering is a Gaussian component to meet the condition to be taken as prospect.Prospect is entered as 255, the back of the body
Scape is entered as 0.Material is thus formed a secondary prospect binary map.Due to containing many noises in prospect binary map, so employing
It is morphologic to open operation by noise reduction to 0, and then rebuild the information due to opening the marginal portion that operation is lost with closed operation.
Eliminate disconnected small noise spot.
Step 2.2.2, the prospect progress self-adaption binaryzation processing extracted to previous step.
Step 2.2.3 and then the projection for carrying out horizontally and vertically direction respectively to binary image.Selected threshold Th 1,
Projected image carries out horizontally and vertically direction traversal respectively.When there is N0< Th, and N0Continuous 10 pixels afterwards are both greater than
Th, then it is assumed that N0For a boundary point of dropper.Obtain after boundary point, intercept dropper target area;Similarly selected threshold Th2,
When there is N1< Th, and N1Preceding 5 points be both greater than Th, rear 5 points are both less than Th, then it is assumed that N1For boundary's point of another side.Similarly
Four boundary points are found, interception target area is template.
Step 2.2.4, with above-mentioned template template matches are carried out, it is consistent with implementing template matching method described in 2.1.2.
Step 2.2.5, the region to above-mentioned matching carry out frame difference and handled.Here the mode once every two frame frames difference is selected
Progress difference, such as 1,4,7,10 ..., this have the advantage that improving computational efficiency.
Step 2.2.6, to difference image, calculation process is opened and closed, then according to 2.1.4 methods describeds carry out pixel value
The statistics of sum.
Flow-speed measurement method is dripped in a kind of above-mentioned intravenous fluid based on machine vision, for the intravenous injection environment back of the body
Scape is complicated, and diversified feature, and when performing step 2, two methods are handled related interference factors, therefore performs
Random selection template matching method is carried out during step 2 or frame difference method is tested the speed.
Therefore, the invention has the advantages that:1st, the present invention devises two kinds of different measuring methods, respectively there is advantage.Make
User can be according to the suitable scheme of different conditions, environmental selection;2nd, the invention provides detailed algorithm model, using setting
For mobile products such as common mobile phones.It is easy to user to understand and practical operation;3rd, instant invention overcomes existing measure skill
The not enough shortcoming that art is complicated, supervision is inconvenient, time-consuming, expensive.Intravenous injection drop is easily measured for more precise and high efficiency
Flow velocity, the workload for mitigating medical worker provides support;4th, shake of the present invention to video has carried out the processing of correlation, it is allowed to clap
It is unstable during taking the photograph, it is not necessary to establishing shot equipment deliberately.Be transplanted to mobile terminal to algorithm, and result accuracy
Bring breakthrough.
Brief description of the drawings
Accompanying drawing 1 is the flow chart of drip-feed flow relocity calculation system.
Accompanying drawing 2 is the algorithm flow chart that video jitter is handled.
Accompanying drawing 3 is that GMM mixed Gauss models extract prospect flow chart.
Accompanying drawing 4 is the algorithm entire flow figure of whole system.
Embodiment
Below by embodiment, and with reference to accompanying drawing, technical scheme is described in further detail.
The invention mainly includes steps
The first step:Obtain Sample video;
Second step:Dropper is considered as moving target.Using GMM mixed Gauss models, foreground target is extracted;
3rd step:Binary conversion treatment;
4th step:Carry out the integral projection in horizontally and vertically direction respectively to previous step bianry image.Threshold value Th is set, point
Other projected image carries out horizontally and vertically direction and traveled through.When there is N0< Th, and N0Continuous 10 pixels afterwards are both greater than Th,
Then think N0For a boundary point of dropper.Obtain after boundary point, intercept dropper target area;
5th step:Template matches are carried out frame by frame, and the region that all frames are matched is dropper position;
Preferably, to try out the measurement under different scenes varying environment, the present invention is entered using two kinds of different idea and methods
Row measurement count.
One kind is template matching method, and one kind is frame difference method.
What template matching method was realized comprises the concrete steps that:
The first step:Hough transformation, finds circular.Extract drop template;
Second step:Template matches are carried out frame by frame, extract matching area;
3rd step:Binaryzation and Morphological scale-space are carried out to matching area;
4th step:Target area pixel value is counted frame by frame;
5th step:Given threshold value Th 2, the frame more than threshold value is thought, containing drop, to record frame number;
6th step:According to the difference for drop frame occur, flow velocity is calculated with the frame per second fps of video.Assuming that continuous two drops video
Frame is respectively N1, N2, then frame number difference is Δ N=N1-N2.As just have under dropping liquid drippage by the time of Δ N frames.Then drip
Fast V=fps/ Δs N.
Second of frame difference method, because the shake of shooting process is influenceed than larger on frame difference, so the present invention uses frame difference method
When, first to carry out stabilization processing.The step that implements of frame difference method is:
The first step:Using the method for the present invention, stabilization processing is carried out;
Second step:Inter-frame difference processing.When not fallen also in dropper due to drop, drop change is slow.Here use
Frame is taken out, every the mode of frame frame difference, detection efficiency is improved.To prevent the phenomenon of Lou frame, according to the dropper mark of different models
Standard, flow velocity is different, and frame period is chosen different;
3rd step:To the image of previous step frame difference, Morphological scale-space is carried out, including opening and closing operation etc..
4th step:Target area pixel value is counted frame by frame;
5th step:Given threshold value Th 3, the frame more than threshold value is thought, containing drop, to record frame number;
6th step:According to the difference for drop frame occur, flow velocity is calculated with the frame per second fps of video.Assuming that continuous two drops frame point
Wei not N1, N2, then frame number difference is Δ N=N1-N2.As just have under dropping liquid drippage by the time of Δ N frames.Then flow velocity V=
fps/ΔN。
Embodiment:
Here is the specific embodiment using the above method.
Such as Fig. 1, shown in 2,3,4, the speed-measuring method of the present embodiment comprises the following steps:
The first step:Sample video is shot, video switchs to still image frame by frame;
Second step:Select speed-measuring method:
2.1st, template matching method
2.1.1, first have to extract drop template, the present invention uses Hough transformation, the method for circular extraction.Drop exists
Form is approximately circular in drip irrigation.Extract after circle, it is (X1, Y1), radius of circle R to choose center of circle O coordinates.Then with coordinate (X1-
R, Y1-R) central point, length and width 2R interception drop templates.
2.1.2, the drop of above onestep extraction is template, and template matches are carried out frame by frame to all frame of video.General template
Algorithm idea:Search pattern T ((m × n) individual pixel) is overlayed and translated on searched figure S (W × H pixel), template is covered
That block region whistle figure S of the searched figure of lidij.I, j are coordinate of the subgraph upper left corner on searched figure S.Hunting zone is:
1≤i≤W-m
1≤j≤W-m
By comparing T and SijSimilitude, complete template matches process.
2.1.3, matching area binaryzation, due to the frame containing drop and the frame without drop difference than larger, so here
Binarization method it is more random.Can be local binarization, global binaryzation can also be self-adaption binaryzation.Binaryzation mesh
Be to preferably carry out the statistics of pixel.
2.1.4, the pixel Data-Statistics Sum of target area.Start from left to right from (0,0), from top to bottom progressively traversing graph
Picture, carries out the cumulative of pixel sum, Jia 1 if being 255 if traversal point pixel value, by that analogy.Calculate the summation that pixel value is 255.
2.2nd, frame difference method
2.2.1, prospect is extracted using GMM mixed Gauss models.First by the average of each Gauss, variance, weights are all set
0 is set to, that is, initializes a model matrix parameter.It is used for training GMM model using the T frames in video.For each pixel,
Set up the GMM model of the maximum GMM_MAX_COMPONT Gauss of its Number of Models.When first pixel is come, it is individually in journey
The initial mean value of its fixation, variance are set in sequence, and weights are set to 1.
In non-first frame training process, when tail pixel value, with the average ratio of above existing Gauss compared with, if
The value of the pixel is with the equal value difference of its model in 3 times of variance, then task belongs to the Gauss.Now carried out more with equation below
Newly:
Whereinα=1/T,
Taken when the value of the pixel and the difference of average be not in the range of its 3 times
After the frame number T for reaching training, the selection of different pixels point GMM numbers adaptively is carried out.It is first
First using right value divided by variance are sorted from big to small to each Gauss, are then chosen the Gauss of foremost B, are met itWherein CfGenerally 0.3
It thus can be very good to eliminate the noise spot in training process.In test phase, to the value and B of pixel of newly arriving
Each average in individual Gauss is compared, if between variance of its difference at 2 times, then it is assumed that be background, otherwise
It is considered prospect.As long as and wherein thering is a Gaussian component to meet the condition to be taken as prospect.Prospect is entered as 255, the back of the body
Scape is entered as 0.Material is thus formed a secondary prospect binary map.Due to containing many noises in prospect binary map, so employing
It is morphologic to open operation by noise reduction to 0, and then rebuild the information due to opening the marginal portion that operation is lost with closed operation.
Eliminate disconnected small noise spot.
2.2.2 self-adaption binaryzation processing, is carried out to the prospect that previous step is extracted.
2.2.3 the projection in horizontally and vertically direction and then to binary image is carried out respectively.Selected threshold Th 1, respectively
Projected image carries out horizontally and vertically direction and traveled through.When there is N0< Th, and N0Continuous 10 pixels afterwards are both greater than Th, then
Think N0For a boundary point of dropper.Obtain after boundary point, intercept dropper target area;Similarly selected threshold Th2, works as appearance
N1< Th, and N1Preceding 5 points be both greater than Th, rear 5 points are both less than Th, then it is assumed that N1For boundary's point of another side.Similarly find four
Individual boundary point, interception target area is template.
2.2.4 template matches, are carried out with above-mentioned template, it is consistent with implementing template matching method described in 2.1.2.
2.2.5 frame difference processing, is carried out to the region of above-mentioned matching.Here carried out from the mode every two frame frames difference once
Difference, such as 1,4,7,10 ..., this have the advantage that improving computational efficiency.
2.2.6, to difference image, calculation process is opened and closed, pixel value sum is then carried out according to 2.1.4 methods describeds
Statistics.
3rd step:Extract the frame number containing drop, set threshold value Th, when a certain frame pixel value for 255 and Sum > Th,
It is containing drop frame then to think this frame.Otherwise it is assumed that drop is not present in this frame.Record the frame number containing drop.Different is quiet
Arteries and veins is injected, and due to the difference of switch controlled number, flow velocity is different.So there is the frame possible more than one of same dropping liquid drop
Frame, it is also possible to have situation about missing.The method that the present invention is used is to record each frame to meet the volume containing drop frame threshold value
Number.Then these numberings are traveled through, when the continuous situation of appearance numbering, then it is assumed that be same frame, calculate the average of serial number
4th step:Flow velocity is calculated according to V=fps/ Δs N.The average frame number of the appearance drop calculated first by above-mentioned stepsThe then adjacent difference of drop frameThe then average of frame-to-frame differencesFor Δ N1, Δ N2, Δ N3... average value.This method for calculating average repeatedly, effectively solves the situation of Lou frame.
Then flow velocityFps is the frame per second of video.
Specific embodiment described herein is only to spirit explanation for example of the invention.Technology neck belonging to of the invention
The technical staff in domain can be made various modifications or supplement to described specific embodiment or be replaced using similar mode
Generation, but without departing from the spiritual of the present invention or surmount scope defined in appended claims.
Claims (3)
1. a kind of intravenous fluid drop flow-speed measurement method based on machine vision, it is characterised in that including:
Step 1, Sample video is shot, video switchs to still image frame by frame;
Step 2, dropper is considered as moving target;Foreground target is extracted, and carries out binary conversion treatment;
Step 3, the frame number containing drop is extracted, threshold value Th is set, when a certain frame pixel value is 255 and Sum > Th, then it is assumed that
This frame is containing drop frame;Otherwise it is assumed that drop is not present in this frame;Record the frame number containing drop;Different vein notes
Penetrate, due to the difference of switch controlled number, flow velocity is different;So there is the more than frame of frame of same dropping liquid drop or having leakage
Situation about falling, records each frame and meets the numbering containing drop frame threshold value;Then these numberings are traveled through, when the continuous feelings of appearance numbering
Condition, then it is assumed that be same frame, calculates the average of serial number
Step 4, flow velocity is calculated according to V=fps/ Δs N;The average frame number of the appearance drop calculated first by above-mentioned stepsThe then adjacent difference of drop frameThe then average of frame-to-frame differencesFor Δ N1, Δ N2, Δ N3... average value;This method for calculating average repeatedly, effectively solves the situation of Lou frame;
Then flow velocityFps is the frame per second of video.
2. a kind of intravenous fluid drop flow-speed measurement method based on machine vision according to claim 1, its feature exists
In, the specific process step of the step 2 is based on two kinds of speed-measuring methods, including:
Speed-measuring method one, template matching method is specifically included:
Step 2.1.1, first have to extract drop template, the present invention uses Hough transformation, the method for circular extraction;Drop exists
Form is approximately circular in drip irrigation;Extract after circle, it is (X1, Y1), radius of circle R to choose center of circle O coordinates;Then with coordinate (X1-
R, Y1-R) central point, length and width 2R interception drop templates;
Step 2.1.2, the drop of above onestep extraction are template, and template matches are carried out frame by frame to all frame of video;General template
Algorithm idea:Search pattern T ((m × n) individual pixel) is overlayed and translated on searched figure S (W × H pixel), template is covered
That block region whistle figure S of the searched figure of lidij;I, j are coordinate of the subgraph upper left corner on searched figure S;Hunting zone is:
1≤i≤W-m
1≤j≤W-m
By comparing T and SijSimilitude, complete template matches process;
Step 2.1.3, matching area binaryzation, due to the frame containing drop and the frame without drop difference than larger, so here
Binarization method it is more random;Can be local binarization, global binaryzation can also be self-adaption binaryzation;Binaryzation mesh
Be to preferably carry out the statistics of pixel;
Step 2.1.4, the pixel Data-Statistics Sum of target area;Start from left to right from (0,0), from top to bottom progressively traversing graph
Picture, carries out the cumulative of pixel sum, Jia 1 if being 255 if traversal point pixel value, by that analogy;Calculate the summation that pixel value is 255;
Speed-measuring method two, frame difference method is specifically included:
Step 2.2.1, utilize GMM mixed Gauss models extract prospect;First by the average of each Gauss, variance, weights are all set
0 is set to, that is, initializes a model matrix parameter;It is used for training GMM model using the T frames in video;For each pixel,
Set up the GMM model of the maximum GMM_MAX_COMPONT Gauss of its Number of Models;When first pixel is come, it is individually in journey
The initial mean value of its fixation, variance are set in sequence, and weights are set to 1;
In non-first frame training process, when tail pixel value, the average ratio with above existing Gauss is compared with if the picture
The value of vegetarian refreshments is with the equal value difference of its model in 3 times of variance, then task belongs to the Gauss;Now it is updated with equation below:
Wherein
Taken when the value of the pixel and the difference of average be not in the range of its 3 timesAfter the frame number T for reaching training, carry out not
With the selection of pixel GMM numbers adaptively;Each Gauss is sorted from big to small with weights divided by variance first, then
The Gauss of foremost B is chosen, makes satisfaction
Wherein CfGenerally 0.3
It thus can be very good to eliminate the noise spot in training process;In test phase, the value and B to pixel of newly arriving are high
Each average in this is compared, if between variance of its difference at 2 times, then it is assumed that be background, otherwise it is assumed that
It is prospect;As long as and wherein thering is a Gaussian component to meet the condition to be taken as prospect;Prospect is entered as 255, and background is assigned
It is worth for 0;Material is thus formed a secondary prospect binary map;Due to containing many noises in prospect binary map, so employing form
That learns opens operation by noise reduction to 0, and then rebuilds the information due to opening the marginal portion that operation is lost with closed operation;Eliminate
Disconnected small noise spot;
Step 2.2.2, the prospect progress self-adaption binaryzation processing extracted to previous step;
Step 2.2.3 and then the projection for carrying out horizontally and vertically direction respectively to binary image;Selected threshold Th 1, respectively
Projected image carries out horizontally and vertically direction and traveled through;When there is N0< Th, and N0Continuous 10 pixels afterwards are both greater than Th, then
Think N0For a boundary point of dropper;Obtain after boundary point, intercept dropper target area;Similarly selected threshold Th2, works as appearance
N1< Th, and N1Preceding 5 points be both greater than Th, rear 5 points are both less than Th, then it is assumed that N1For boundary's point of another side;Similarly find four
Individual boundary point, interception target area is template;
Step 2.2.4, with above-mentioned template template matches are carried out, it is consistent with implementing template matching method described in 2.1.2;
Step 2.2.5, the region to above-mentioned matching carry out frame difference and handled;Here carried out from the mode every two frame frames difference once
Difference, such as 1,4,7,10 ..., this have the advantage that improving computational efficiency;
Step 2.2.6, to difference image, calculation process is opened and closed, pixel value sum is then carried out according to 2.1.4 methods describeds
Statistics.
3. a kind of intravenous fluid drop flow-speed measurement method based on machine vision according to claim 2, its feature exists
In, it is complicated for intravenous injection environmental background, and diversified feature, when performing step 2, two methods to relevant interference because
Element is handled, therefore progress random selection template matching method or frame difference method are tested the speed when performing step 2.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710178754.XA CN106997587B (en) | 2017-03-23 | 2017-03-23 | Method for measuring flow velocity of intravenous injection liquid drops based on machine vision |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710178754.XA CN106997587B (en) | 2017-03-23 | 2017-03-23 | Method for measuring flow velocity of intravenous injection liquid drops based on machine vision |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106997587A true CN106997587A (en) | 2017-08-01 |
CN106997587B CN106997587B (en) | 2020-06-23 |
Family
ID=59431870
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710178754.XA Expired - Fee Related CN106997587B (en) | 2017-03-23 | 2017-03-23 | Method for measuring flow velocity of intravenous injection liquid drops based on machine vision |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106997587B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112986105A (en) * | 2021-02-07 | 2021-06-18 | 睿科集团(厦门)股份有限公司 | Liquid drop counting and speed measuring method based on machine vision |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103179995A (en) * | 2010-07-15 | 2013-06-26 | 陶锴 | Iv monitoring by video and image processing |
US20150350608A1 (en) * | 2014-05-30 | 2015-12-03 | Placemeter Inc. | System and method for activity monitoring using video data |
CN105498042A (en) * | 2016-01-08 | 2016-04-20 | 山东师范大学 | Video-based non-light-shielding type transfusion automatic alarm method and device thereof |
CN105664297A (en) * | 2016-03-14 | 2016-06-15 | 英华达(南京)科技有限公司 | Infusion monitoring method, system and device |
-
2017
- 2017-03-23 CN CN201710178754.XA patent/CN106997587B/en not_active Expired - Fee Related
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103179995A (en) * | 2010-07-15 | 2013-06-26 | 陶锴 | Iv monitoring by video and image processing |
US20150350608A1 (en) * | 2014-05-30 | 2015-12-03 | Placemeter Inc. | System and method for activity monitoring using video data |
CN105498042A (en) * | 2016-01-08 | 2016-04-20 | 山东师范大学 | Video-based non-light-shielding type transfusion automatic alarm method and device thereof |
CN105664297A (en) * | 2016-03-14 | 2016-06-15 | 英华达(南京)科技有限公司 | Infusion monitoring method, system and device |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112986105A (en) * | 2021-02-07 | 2021-06-18 | 睿科集团(厦门)股份有限公司 | Liquid drop counting and speed measuring method based on machine vision |
Also Published As
Publication number | Publication date |
---|---|
CN106997587B (en) | 2020-06-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Datta et al. | Person-on-person violence detection in video data | |
CN106096577B (en) | A kind of target tracking method in camera distribution map | |
Siagian et al. | Biologically inspired mobile robot vision localization | |
CN102175693B (en) | Machine vision detection method of visual foreign matters in medical medicament | |
CN106131469B (en) | Ball intelligent robot coach and judgment system based on machine vision | |
CN109271888A (en) | Personal identification method, device, electronic equipment based on gait | |
CN107886503A (en) | A kind of alimentary canal anatomical position recognition methods and device | |
CN109522854A (en) | A kind of pedestrian traffic statistical method based on deep learning and multiple target tracking | |
CN106407935A (en) | Psychological test method based on face images and eye movement fixation information | |
CN104224129B (en) | A kind of vein blood vessel depth recognition method and prompt system | |
WO2020042542A1 (en) | Method and apparatus for acquiring eye movement control calibration data | |
CN103295016A (en) | Behavior recognition method based on depth and RGB information and multi-scale and multidirectional rank and level characteristics | |
CN107871120A (en) | Competitive sports based on machine learning understand system and method | |
CN103413120A (en) | Tracking method based on integral and partial recognition of object | |
CN103325115B (en) | A kind of method of monitoring people counting based on overhead camera head | |
CN107408119A (en) | Image retrieving apparatus, system and method | |
CN108335747A (en) | Cognitive training system | |
CN104615997B (en) | A kind of face method for anti-counterfeit based on multiple-camera | |
CN103593679A (en) | Visual human-hand tracking method based on online machine learning | |
CN108090485A (en) | Display foreground extraction method based on various visual angles fusion | |
CN108009468A (en) | A kind of marathon race anti-cheat method, electronic equipment and storage medium | |
CN104036231A (en) | Human-body trunk identification device and method, and terminal-point image detection method and device | |
Błażek et al. | An unorthodox view on the problem of tracking facial expressions | |
CN104778676A (en) | Depth ranging-based moving target detection method and system | |
CN110287907A (en) | A kind of method for checking object and device |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20200623 Termination date: 20210323 |
|
CF01 | Termination of patent right due to non-payment of annual fee |