CN102073070A - Freezing detection method based on image processing - Google Patents
Freezing detection method based on image processing Download PDFInfo
- Publication number
- CN102073070A CN102073070A CN2011100065498A CN201110006549A CN102073070A CN 102073070 A CN102073070 A CN 102073070A CN 2011100065498 A CN2011100065498 A CN 2011100065498A CN 201110006549 A CN201110006549 A CN 201110006549A CN 102073070 A CN102073070 A CN 102073070A
- Authority
- CN
- China
- Prior art keywords
- disk
- image
- disc
- sequence image
- detection method
- 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
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Image Analysis (AREA)
Abstract
The present invention discloses a freezing detection method based on image processing, which can realize unattended and automated observation of a freezing phenomenon. The method comprises the following steps: making different character marks on the surface of a disc, arranging the disc into a water tank after marking, exerting instantaneous force on the disc according to a certain time interval, and collecting a sequence image of the disc; and matching the shape and the position of the last frame of the sequence image with the shapes and the positions of preceding frames character by character. If the matching is successful, the disc does not rotate during collection of the sequence image, and water is frozen. The rotation of the disc can be detected automatically by adopting the image processing algorithm, and the water freezing can be detected indirectly according to the results. No human intervention is needed during detect, the operation is simple and the cost is low.
Description
Technical field
The invention belongs to Flame Image Process and meteorological observation field, be specifically related to a kind of mode and come the indirect observation water surface whether the phenomenon of freezing takes place by Flame Image Process.
Background technology
Meteorological observation is the basis of weather service work.Surface weather observation is the important component part of meteorological observation, it is that meteorological condition in the earth surface certain limit and change procedure thereof are carried out systematically, observe continuously and measure, for weather forecast, meteorologic information, climatic analysis, scientific research and Meteorological Services provide important basis.
Be defined as for " freezing " in meteorological industry standard " surface weather observation standard " the 4th part (QX/T48-2007) the weather phenomenon observation of the People's Republic of China (PRC): " the surface of open water (water that comprises evaporator) glaciation." simultaneously, also spell out in the points for attention for observation and record, " observation person on duty should observe and write down the whole weather phenomena that appear in the vision area at any time.Night is the weather station of the class of keeping not, to the weather phenomenon that occur night, should judge record as far as possible." as seen, the phenomenon of freezing can only rely on the method for artificial observation to judge and record as a basic ground phenomena at present.Though artificial observation is comparatively accurate for the differentiation of phenomenon itself, have following insoluble problem: 1, most of phenomenon of freezing occurs in night, and website and observation person itself have all been proposed certain requirement.2, the moment of taking place for the phenomenon of freezing is difficult to record accurately.
Summary of the invention
The present invention proposes a kind of icing phenomenon robotization recognition methods, can on the basis of minimum maintenance cost, realize the automatic observation of icing phenomenon based on Flame Image Process.
A kind of icing detection method based on Flame Image Process is specially: make the distinct symbols mark in a disc surfaces, be placed on behind the mark in the tank, at interval disk is applied transient force by certain hour, gather the sequence image of disk simultaneously; With the last frame of sequence image pursues symbol respectively with each two field picture of front shape and location matches, if all successes of shape and location matches show that disk does not rotate in the acquisition time section of sequence image, then the water surface freezes.
As preferred implementation, described shape and location matches mode by symbol is specially:
(1) in each two field picture of sequence image, extracts the area-of-interest that comprises disk;
(2) connected domain and the disc area of each symbol of extraction in area-of-interest keep the symbol connected domain that is positioned at disc area;
(3) last frame of sequence image is done one by one shape and location matches with the stet connected domain of each two field picture of front respectively.
As preferred implementation, described area-of-interest is the boundary rectangle of disk.
As preferred implementation, disk is applied contactless transient force.
The present invention also provides another kind of icing detection method based on Flame Image Process, is specially: a disk is placed tank, at interval disk is applied transient force by certain hour, gather the sequence image of disk simultaneously; The last frame of sequence image is carried out the similarity coupling with each two field picture of front respectively, if all successes of similarity coupling show that disk does not rotate in the acquisition time section of sequence image, then the water surface freezes.
Technique effect of the present invention is embodied in: after on the image disc area being carried out disposable mark, utilize the automatic test disks of image processing algorithm whether to take place to rotatablely move, and by this as a result the indirect detection water surface whether icing phenomenon has taken place, in testing process, this scheme has not to be had under the artificial situation of intervening, realization is to the automatic observation of the phenomenon of freezing, and can provide the moment that the phenomenon of freezing takes place.Simultaneously, this scheme equipment needed thereby is simple, and cost is lower.
Description of drawings
Fig. 1 comes the disk legend of symbol for mark;
Fig. 2 is the image processing algorithm overview flow chart;
Fig. 3 is the zone diagram of user's mark;
Fig. 4 is an image pre-treatment step process flow diagram;
Fig. 5 extracts and the filtration step process flow diagram for the symbol connected domain;
Fig. 6 is the Symbol recognition flow chart of steps;
Fig. 7 differentiates process flow diagram for the phenomenon of freezing.
Embodiment
Below in conjunction with accompanying drawing preferred embodiment of the present invention is described.
Object of observation of the present invention is the circle evaporation tank that outdoor placement is filled with water.Place a light plastic disk at the water surface, mark arabic numeral on the disk, as 1-9, the capitalization English letter is as characters such as A-T, as shown in Figure 1.At micro-air jet system of disk overhung.This device continues jet fixedly duration every the set time, and each jet disk that all can make rotates motion.As a kind of drive unit, adopt the benefit of jet mode to be not contact with icing disk or water surface generation physics, the simultaneously of short duration jet duration also can the condition that disk can slowly rotate when guaranteeing not have to take place to freeze under, reduce the interference that this spontaneous phenomenon generation of freezing is caused as much as possible.Utilize camera timing acquiring tank fixed area image.If the water surface does not freeze, in the interior at certain time intervals image sequence of gathering, the character on the disk can change, if the water surface takes place to freeze, disk is with frozen, and in the interior at certain time intervals image sequence of gathering, the character on the disk can remain unchanged.By above-mentioned two kinds of situations, can judge whether the water surface phenomenon of freezing takes place.
Following explanation will suppose that we have obtained outdoor freeze water slot part area image sequence, and the image sequence acquisition time interval is 10 minutes.
Image processing algorithm whole detection flow process describes in detail below and uses detection step of the present invention as shown in Figure 2:
(1) area-of-interest markers step
This flow chart of steps as shown in Figure 3, to any t two field picture of outdoor freeze water slot part area image sequence I
t, the manual markings image I
tMiddle disk outward flange.The mode of mark is that (x y) as starting point, draws closed curve along external periphery outline, and point on the curve and internal point compositing area point are gathered A, and A should comprise visible integrated disc portions on the image fully from any one point of disk outward flange.Simultaneously, ask for following four numerical value in the region point set A:
Wherein, (x y) is any point among the regional A to p.
, be designated as area-of-interest with these 4 rectangular areas that are worth generation:
Rect(p(x
Min,y
Min),p(x
Max,y
Max))
P (x wherein
Min, y
Min) be the upper left corner coordinate of rectangle, p (x
Max, y
Max) be the lower right corner coordinate of rectangle.
Generate secondary 8 gray scale thumbnails as the surveyed area template, its width and highly be equal to the width in Rect zone and highly, this image be designated as Mask (x, y), the gray-scale value of each pixel is according to following regular value:
Wherein, Mask (x, y) be in the image coordinate for (x y) locates the gray-scale value of pixel.The purpose that generates this template image is in order further to dwindle sensing range in area-of-interest, to filter and disturb, can using in subsequent step.
(2) image pre-treatment step
Image pre-treatment step flow process as shown in Figure 4.The image that success is each time gathered carries out following processing:
1. 4 apex coordinates according to the Rect zone in the step (1) carry out cutting to original image;
2. image after the cutting is carried out gray processing;
3. adopt maximum between-cluster variance criterion (OTSU) (Nobuyuki Otsu.A ThresholdSelection Method from Gray-Level Histograms, IEEE Trans.Systems, Man, and Cybernetics, 9 (1): 62~66,1979.) carry out the self-adaptation dividing processing;
4. (x y) carries out the minimum value interaction process to the middle surveyed area template Mask that generates in segmentation result and (1);
The result who supposes the 2nd gray processing of this step is Gray (x, y), the 3rd middle segmentation result image be Binary (x, y), the operation purpose that this step is the 4th is to remove to fall among the area-of-interest, but is in segmentation result outside the disc area to the influence of mark connected domain.Its result image I ntersect (x, y) be defined as segmentation result image B inary (x, y) and (1) in template image Mask (x, y) the correspondence position grey scale pixel value is got the smaller.
Intersect(x,y)=Min(Binary(x,y),Mask(x,y))
(3) the character connected domain is extracted and filtration step
Extraction of character connected domain and filtration step flow process are as shown in Figure 5.The purpose of this step is to extract the image of each character, and interference is filtered.(x y) carries out connected component labeling, extracts all connected domains on the image, supposes that the connected domain set that certain frame extracts is B, Blob to mutual back image I ntersect to utilize the connected component labeling algorithm
kBe k connected domain wherein, Blob
k∈ B, bw, bh are Blob
kBoundary rectangle is the width and the height of unit with the pixel, according to following formula connected domain is filtered.
Wherein, W
Min, W
Max, H
Min, H
MaxBe respectively the minimum widith of predefined character-circumscribed rectangle, breadth extreme, minimum constructive height and maximum height.Dispose is the sign that whether abandons this connected domain, is 1 to abandon, and be 0 reservation.Because the actual size of character all is known on camera antenna height and the disk, and be changeless, so the resolution of image is changeless in testing process, promptly the character boundary of Xian Shiing is known to changeless, can set W according to these data
Min, W
Max, H
Min, H
Max4 threshold values.Unless the purpose of this step is by the restriction of the connected domain size character connected domain of making a return journey.
If k the character connected domain that present frame extracts is Blob
k, its marginal point set is Edge
k(x, y).(x y) adopts single order gradient Sobel operator to calculate Edge to the gray processing result images Gray in the step (2)
k(x, y) in each point at x direction and y direction gradient, and calculate its mould value, be designated as g
k(x, y),
Wherein, I
x(x y) is the gradient image of x direction, I
y(x y) is the gradient image of y direction.
Calculate g
k(x, average y) and variance:
According to all connected domains of following rule-based filtering:
In the practical application, M is the average threshold value, and span is [160,200], and D is a variance threshold values, and span is [80,100].Dispose is the sign that whether abandons this connected domain, is 1 to abandon, and be 0 reservation.
(4) character recognition step
The character recognition step flow process as shown in Figure 6, this flow process is at the character recognition of each character connected domain.Extract character connected domain proper vector by Fourier descriptors under the employing polar coordinates, and adopt the classification that K nearest neighbor classifier (KNN) is judged this proper vector.
1. establishing k character connected domain of certain frame image is BlobImage
k(x y), carry out logicization to gray-scale value by the following method, and calculates the quality mass and the center-of-mass coordinate (x of normalized image
c, y
c).
Wherein, (x y) is the result of k character connected domain image logicization to f, and M is the height of k character connected domain image, and N is the width of k character connected domain image.
2. with (x
c, y
c) be that polar coordinate system, computed image f (x, maximum radius R y) are constructed in the center of circle
Max, and to f (x y) carries out Fourier transform under the polar coordinates, obtain real part coefficient fr after the conversion (u, v), and imaginary part coefficient fi (u, v).Computing formula is as follows:
Wherein, cos, sin, atan are respectively cosine function, sine function, arctan function, generate k character connected domain proper vector FD
k
Wherein, m is called radius resolution, and span is [4,10], and n is called angular resolution, and span is [4,20], FD
kLength be m * n.
Before handling in real time, the character of selecting to occur from real image is as sample, sample must cover the character that might occur, each character becomes a classification, selects a plurality of identical characters images as sample in each classification, sample set is designated as:
Wherein,
Represent n sample in the m class, for example in the following formula
The K that represents the 1st class character
1Individual sample,
The K that represents the 2nd class character
2Individual sample, in the practical application, the number of samples of every class character should be more than or equal to 5.
Each sample is calculated its eigenwert according to the method for step (4), and it is kept at this locality, when handling in real time, adopt each character to carry out online eigenvalue calculation, and utilize K nearest neighbor classifier (KNN) to judge the affiliated classification of character of this appearance occurring.In the practical application, K value 3 or 5.
(5) icing phenomenon discriminating step
The phenomenon discriminating step flow process of freezing as shown in Figure 7.Come decision making package whether the phenomenon of freezing takes place according to following two criterions.
Whether criterion 1. changes identification according to character
When the image sequence of gathering surpasses N
1In the time of frame, establish present frame and be numbered t, k>N
1, the character set that identifies is:
CharList
t={char
t 1,char
t 2,...,char
t i,...,char
t n},i=0,1,...n
The character connected domain boundary rectangle centre coordinate set that present frame identifies is:
Wherein, char
t iFor i character of present frame order appearance, according to series arrangement from left to right,
Be char
t iCharacter connected domain boundary rectangle centre coordinate.
If, then be judged to be icing phenomenon taken place if following formula is set up.
Wherein, Δ t=1,2 ..., N
1, N
1Be maximum reference frame number; SR is an image spatial resolution, is a constant after video camera and disk relative position are fixing, and unit be millimeter that MaxCount is the number of pixels that sets in advance, MaxCount and N in the practical application
1Value should satisfy following formula:
Wherein, interval is the two frame period times of front and back, is a constant in testing process, and h is hour that 2h promptly represents 2 hours.
If take place to freeze, then export (t-N
1) time of taking place for the phenomenon of freezing image frame grabber time, flow process finishes.
If do not take place to freeze, then change criterion 2 over to.
Whether criterion 2. changes identification according to related coefficient
When the image sequence of gathering surpasses N
2In the time of frame, begin to calculate related coefficient g
tThe value of (Δ t), g
t(Δ t) is defined as:
Wherein, t is the numbering of present frame, t>N
2, Δ t={1,2, L, N
2, should satisfy following formula in the practical application:
Wherein, interval is the two frame period times of front and back, and h is hour that 2.5h was 2 hours 30 minutes.A is the surveyed area Rect (p (x of definition in (1)
Min, y
Min), p (x
Max, y
Max)), S (A) is the area of A, computing formula is:
S(A)=|(x
Max-x
Min)×(y
Max-y
Min)|
If satisfy following formula, then be judged to the phenomenon of freezing and take place:
min(g
t(Δt))>Th,Δt={1,2,L,N
2}
Wherein, Th is minimum similarity threshold, and in the practical application, span is [0.95,0.99].
If take place to freeze, then export (t-N
2) time of taking place for the phenomenon of freezing image frame grabber time, flow process finishes.
Claims (6)
1. the icing detection method based on Flame Image Process is specially: make the distinct symbols mark in a disc surfaces, be placed on behind the mark in the tank, at interval disk is applied transient force by certain hour, gather the sequence image of disk simultaneously; With the last frame of sequence image pursues symbol respectively with each two field picture of front shape and location matches, if all successes of shape and location matches show that disk does not rotate in the acquisition time section of sequence image, then the water surface freezes.
2. a kind of icing detection method based on Flame Image Process according to claim 1 is characterized in that, described shape and location matches mode by symbol is specially:
(1) in each two field picture of sequence image, extracts the area-of-interest that comprises disk;
(2) connected domain and the disc area of each symbol of extraction in area-of-interest keep the symbol connected domain that is positioned at disc area;
(3) last frame of sequence image is done one by one shape and location matches with the stet connected domain of each two field picture of front respectively.
3. a kind of icing detection method based on Flame Image Process according to claim 2 is characterized in that described area-of-interest is the boundary rectangle of disk.
4. according to claim 1 or 2 or 3 described a kind of icing detection methods, it is characterized in that, disk is applied contactless transient force based on Flame Image Process.
5. the icing detection method based on Flame Image Process is specially: a disk is placed tank, at interval disk is applied transient force by certain hour, gather the sequence image of disk simultaneously; The last frame of sequence image is carried out the similarity coupling with each two field picture of front respectively, if all successes of similarity coupling show that disk does not rotate in the acquisition time section of sequence image, then the water surface freezes.
6. a kind of icing detection method based on Flame Image Process according to claim 5 is characterized in that, disk is applied contactless transient force.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2011100065498A CN102073070A (en) | 2011-01-13 | 2011-01-13 | Freezing detection method based on image processing |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2011100065498A CN102073070A (en) | 2011-01-13 | 2011-01-13 | Freezing detection method based on image processing |
Publications (1)
Publication Number | Publication Date |
---|---|
CN102073070A true CN102073070A (en) | 2011-05-25 |
Family
ID=44031685
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2011100065498A Pending CN102073070A (en) | 2011-01-13 | 2011-01-13 | Freezing detection method based on image processing |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN102073070A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110398998A (en) * | 2019-05-06 | 2019-11-01 | 李娜 | Computer heating control platform based on temperature difference detection |
CN108257085B (en) * | 2017-08-24 | 2021-07-30 | 北京航空航天大学 | Mechanical deicing process detection method based on image processing technology |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101893724A (en) * | 2010-08-16 | 2010-11-24 | 中国气象局气象探测中心 | Automatic freezing observation method and device for ground meteorological observation |
-
2011
- 2011-01-13 CN CN2011100065498A patent/CN102073070A/en active Pending
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101893724A (en) * | 2010-08-16 | 2010-11-24 | 中国气象局气象探测中心 | Automatic freezing observation method and device for ground meteorological observation |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108257085B (en) * | 2017-08-24 | 2021-07-30 | 北京航空航天大学 | Mechanical deicing process detection method based on image processing technology |
CN110398998A (en) * | 2019-05-06 | 2019-11-01 | 李娜 | Computer heating control platform based on temperature difference detection |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Guan et al. | Robust traffic-sign detection and classification using mobile LiDAR data with digital images | |
CN101807352B (en) | Method for detecting parking stalls on basis of fuzzy pattern recognition | |
CN101430195B (en) | Method for computing electric power line ice-covering thickness by using video image processing technology | |
CN105760831B (en) | It is a kind of to be taken photo by plane the pedestrian tracting method of infrared video based on low latitude | |
CN108921201B (en) | Dam defect identification and classification method based on feature combination with CNN | |
CN107480585A (en) | Object detection method based on DPM algorithms | |
CN101577812A (en) | Method and system for post monitoring | |
CN103605971B (en) | Method and device for capturing face images | |
CN104091175B (en) | A kind of insect automatic distinguishing method for image based on Kinect depth information acquiring technology | |
CN105574488A (en) | Low-altitude aerial infrared image based pedestrian detection method | |
CN103164858A (en) | Adhered crowd segmenting and tracking methods based on superpixel and graph model | |
CN101697007A (en) | Radar image-based flyer target identifying and tracking method | |
CN105243386A (en) | Face living judgment method and system | |
CN110197231A (en) | The bird feelings detecting devices merged based on visible light and infrared light image and recognition methods | |
CN102289686A (en) | Method for identifying classes of moving targets based on transfer learning | |
CN103279760B (en) | A kind of plant quarantine larva real-time grading method | |
An et al. | An automated airplane detection system for large panchromatic image with high spatial resolution | |
CN102622584A (en) | Method for detecting mask faces in video monitor | |
Kwok | Declassified high-resolution visible imagery for Arctic sea ice investigations: An overview | |
CN106156695B (en) | Outlet and/or entrance area recognition methods and device | |
CN104951799A (en) | SAR remote-sensing image oil spilling detection and identification method | |
CN109190624A (en) | Kitchen fume concentration detection method based on image procossing | |
CN106815583A (en) | A kind of vehicle at night license plate locating method being combined based on MSER and SWT | |
CN102663411A (en) | Recognition method for target human body | |
CN110288623B (en) | Data compression method for unmanned aerial vehicle maritime net cage culture inspection image |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
WD01 | Invention patent application deemed withdrawn after publication |
Open date: 20110525 |