CN109359646A - Liquid level type Meter recognition method based on crusing robot - Google Patents
Liquid level type Meter recognition method based on crusing robot Download PDFInfo
- Publication number
- CN109359646A CN109359646A CN201811150848.7A CN201811150848A CN109359646A CN 109359646 A CN109359646 A CN 109359646A CN 201811150848 A CN201811150848 A CN 201811150848A CN 109359646 A CN109359646 A CN 109359646A
- Authority
- CN
- China
- Prior art keywords
- region
- instrument
- liquid level
- oil level
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/22—Image preprocessing by selection of a specific region containing or referencing a pattern; Locating or processing of specific regions to guide the detection or recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/02—Recognising information on displays, dials, clocks
Abstract
The liquid level type Meter recognition method based on crusing robot that the invention discloses a kind of chooses instrument image placed in the middle as template image using Instrument image data set training Adaboost classifier, and for each inspection point；The instrument picture to be detected for obtaining specified inspection point carries out coarse positioning and accurate positioning in conjunction with Adaboost classifier and template image, filters out dial plate region；According to the color difference in oil level region and background area, from dial plate extracted region oil level region；According to shape, luminance difference, from oil level extracted region oil level indicator object block, liquid level type instrument registration is calculated according to oil level indicator region bound and object block position.The present invention utilizes machine learning, can detect the liquid level type instrument registration identified under a variety of illumination, attitudes vibration.
Description
Technical field
The present invention relates to electric inspection process robot fields, and in particular to a kind of liquid level type instrument knowledge based on crusing robot
Other method.
Background technique
Electric inspection process robot needs to realize autonomous localization and navigation in substation, the identification of field instrument registration, fills automatically
The basic functions such as electricity, wherein the instrument and meter registration for detecting live power equipment is the most crucial function of electric inspection process robot.
Most of liquid level type instrument does not have the transporting function of intelligence instrument due to cost and history, can only utilize inspection
Robot goes to read instrument registration by the method for computer vision.Want to accurately identify liquid level type instrument registration, it is necessary to quasi-
Really detect the position of liquid level type instrument in visual pattern.For most of liquid level type instrument in outdoor, illumination condition is complicated, patrols simultaneously
It is also different to examine robot shooting angle, when being detected and identified using traditional image processing means, the registration of detection is unstable
Determine, and the deviation of practical registration is also larger.
Summary of the invention
The liquid level type Meter recognition method based on crusing robot that the purpose of the present invention is to provide a kind of, improves not
The stability and precision identified with registration under the conditions of illumination difference posture.
The technical solution for realizing the aim of the invention is as follows: a kind of liquid level type Meter recognition side based on crusing robot
Method includes the following steps:
Step 1, classifier training: using Instrument image data set training Adaboost classifier, and being each inspection point
An instrument image placed in the middle is chosen as template image；
Step 2, instrument zone location: the instrument picture to be detected of specified inspection point is obtained, in conjunction with Adaboost classifier
Coarse positioning and accurate positioning are carried out with template image, filters out dial plate region；
Step 3 extracts oil level region: according to the color difference in oil level region and background area, from dial plate extracted region oil
Position region；
Step 4, detector indicate number: according to shape, luminance difference, from oil level extracted region oil level indicator object block, according to oil
Position meter region bound and object block position calculate liquid level type instrument registration.
Compared with prior art, the present invention its remarkable advantage are as follows: 1) present invention has merged robot localization information, utilizes machine
The existing location information of device people, so that position multiplicity is high (such as the positioning for being lower than 5cm), the variations such as scale, rotation are smaller,
The problem of carrying out rough detection using phase correlation on the basis of this, reducing missing inspection to greatest extent, substantially all tables can be examined；
2) present invention utilizes machine learning, can detect the instrument under a variety of illumination, attitudes vibration；3) the invention detects that target image
Afterwards, it is operated by opening operation and Otsu algorithm etc., solves the registration identification of the liquid level type instrument under the conditions of different illumination, posture
Problem.
Detailed description of the invention
Fig. 1 is that the present invention is based on the flow charts of the liquid level type Meter recognition method of crusing robot.
Specific embodiment
In the following with reference to the drawings and specific embodiments, the present invention program is further illustrated.
As shown in Figure 1, the liquid level type Meter recognition method based on crusing robot, includes the following steps:
Step 1, classifier training: using Instrument image data set training Adaboost classifier, and being each inspection point
An instrument image placed in the middle is chosen as template image.Since Instrument image data set contains different illumination, posture condition
Under Instrument image, how subsequent instrument detection environment to change, can accurately extract.
Step 2, instrument zone location: the instrument picture to be detected of specified inspection point is obtained, in conjunction with Adaboost classifier
Coarse positioning and accurate positioning are carried out with template image, filters out dial plate region, method particularly includes:
Step 2.1 treats the target instrument detected in instrument picture using plum forests Fourier transformation and phase coherent techniques
Region carries out coarse positioning, obtains coarse positioning target gauge field；
Step 2.2, using trained Adaboost classifier treat detection instrument picture be accurately positioned, obtain several
Object candidate area；
Step 2.3, the perceptual hash, mutual information and the friendship that calculate each object candidate area and three kinds of parameter indexes of ratio, do
Weighting finds out the confidence level of each object candidate area, determines alternative testing result；
Calculate the friendship of each object candidate area and coarse positioning target instrument region and than parameter IOU, formula are as follows:
In formula, C is coarse positioning target instrument region, niFor object candidate area.
The perceptual hash index of each object candidate area and template image is calculated, method particularly includes: by target candidate area
Area image and template image zoom to same size, carry out cosine transform, the low frequency in the image upper left corner after choosing cosine transform
Region removes the DC component of coordinate (0,0), obtains a feature vector, calculates object candidate area image and template image
Feature vector Hamming distance, perceptually Hash index.
Calculate the mutual information index of each object candidate area image and template image, formula are as follows:
In formula, G(X)、H(Y)The respectively number of template image and candidate image gray-scale pixels, W, H are respectively candidate region
Image is wide, high.
It does weighting by the friendship of each object candidate area and than three kinds of IOU, mutual information index and perceptual hash indexs and asks
The confidence level of each object candidate area out, formula are as follows:
Confidence=1- (pHash+1/I (G(X), H(Y)))/(IOU+D)
In formula, I (G(X), H(Y)) it is mutual information index, pHash is perceptual hash index, and for IOU to hand over and than index, D is to set
Fixed constant, by the maximum object candidate area of confidence level alternately testing result.
Step 2.4 screens final target instrument region, i.e. dial plate region according to the index value of alternative testing result, if
The IOU of alternative testing result meets while being less than given threshold thresholdIOU, and (pHash+1/I (G(x), H(Y))) be greater than
When threshold value thresholdA, using coarse positioning target instrument region as final goal instrument region, otherwise in case selecting testing result
As final goal instrument region.Usually setting threshold value thresholdIOU value range 0.1~0.4, threshold value thresholdA
Value range 10~50.
Step 3 extracts oil level region: since oil level region is different with the color of background area (dark and light color), conversion
To have notable difference after gray scale, oil level region, specific steps can be extracted according to this feature are as follows:
Step 3.1 carries out gray processing processing to dial plate image, and in order to reduce, illumination, dust are to the shadow of picture gray scale etc.
It rings, reuses histogram equalization, gaussian filtering removal interference；
Step 3.2, due to object it is brighter than background, small lump can be excluded by carrying out opening operation herein, and deletion cannot
Subject area comprising structural element, the profile of smooth object disconnect narrow connection, remove tiny protrusion.Open fortune
At last by first realizing to Image erosion reflation, Principle representation formula is as follows:
Dst=open (src, element)=dilate (erode (src, element))
Step 3.3, using Otsu algorithm to dial plate region binaryzation.Otsu algorithm principle is as follows:
For image I (x, y), if the size of image is M × N, the segmentation threshold of prospect (i.e. target) and background is T, general
The overall average gray scale of image is denoted as μ, and inter-class variance is denoted as g, and since the background of image is darker, the gray value of pixel is less than in image
The pixel of threshold value T is background, and the number of pixels of background is denoted as N0, background pixel points account for the ratio of entire image and are denoted as ω1,
Average gray is denoted as μ1；Pixel of the pixel grey scale greater than threshold value T is prospect, and the number of pixels of target is denoted as N1, belong to the picture of prospect
The ratio that vegetarian refreshments number accounts for entire image is denoted as ω0, average gray is denoted as μ0, then have:
ω0=N0/M×N
ω1=N1/M×N
N0+N1=M=N
ω0+ω11=1
μ=ω0*μ0+ω1*μ1
G=ω0(μ0-μ)^2+ω1(μ1-μ)^2
Obtain equivalence formula:
G=ω0ω1(μ0-μ1)^2
It obtains making the maximum threshold value T of inter-class variance g, as required segmentation threshold using the method for traversal.
Step 3.4 carries out subregion opening operation, first carries out opening operation to oil level pipe or so region, removes left and right noise；Again
Opening operation is carried out to lower regions, removes noise up and down；
Step 3.5, according to grey value difference, selected Threshold segmentation goes out oil level region, obtains oil level indicator region.It is normally set up
Threshold range is 40-70.
Step 4, detector indicate number: according to shape, luminance difference, using contours segmentation extraction oil level indicator object block, and by
Oil level indicator region bound and object block position calculate liquid level type instrument registration, specific steps are as follows:
Step 4.1 carries out profile scan to oil level indicator area image using Binarization methods, extracts largest connected domain；
Step 4.2 finds out dial plate intermediate region, and opening operation eliminates object block ambient noise, obtains object block position；
Step 4.3, according to the object block region divided, liquid level type is calculated by oil level indicator region bound and object block position
Instrument registration.
Claims (8)
1. a kind of liquid level type Meter recognition method based on crusing robot, which comprises the steps of:
Step 1, classifier training: it using Instrument image data set training Adaboost classifier, and is chosen for each inspection point
One instrument image placed in the middle is as template image；
Step 2, instrument zone location: the instrument picture to be detected of specified inspection point is obtained, in conjunction with Adaboost classifier and mould
Plate image carries out coarse positioning and accurate positioning, filters out dial plate region；
Step 3 extracts oil level region: according to the color difference in oil level region and background area, from dial plate extracted region oil level area
Domain；
Step 4, detector indicate number: according to shape, luminance difference, from oil level extracted region oil level indicator object block, according to oil level indicator
Region bound and object block position calculate liquid level type instrument registration.
2. the liquid level type Meter recognition method according to claim 1 based on crusing robot, which is characterized in that step 2
In, screening dial plate region method particularly includes:
Step 2.1 treats the target instrument region detected in instrument picture using plum forests Fourier transformation and phase coherent techniques
Coarse positioning is carried out, coarse positioning target gauge field is obtained；
Step 2.2, using trained Adaboost classifier treat detection instrument picture be accurately positioned, obtain several targets
Candidate region；
Step 2.3, the perceptual hash, mutual information and the friendship that calculate each object candidate area and three kinds of parameter indexes of ratio, weight
The confidence level for finding out each object candidate area determines alternative testing result；
Step 2.4, according to the index value of alternative testing result, selected from coarse positioning target instrument region and alternative testing result
Final goal instrument region, i.e. dial plate region.
3. the liquid level type Meter recognition method according to claim 2 based on crusing robot, which is characterized in that step
In 2.3, alternative testing result is determined method particularly includes:
Calculate the friendship of each object candidate area and coarse positioning target instrument region and than parameter IOU, formula are as follows:
In formula, C is coarse positioning target instrument region, niFor object candidate area；
The perceptual hash index of each object candidate area and template image is calculated, method particularly includes: by object candidate area figure
Picture zooms to same size with template image, carries out cosine transform, the low frequency region in the image upper left corner after choosing cosine transform,
The DC component for removing coordinate (0,0) obtains a feature vector, calculates the feature of object candidate area image and template image
The Hamming distance of vector, perceptually Hash index；
Calculate the mutual information index of each object candidate area image and template image, formula are as follows:
In formula, G(X)、H(Y)The respectively number of template image and candidate image gray-scale pixels, W, H are respectively candidate region image
It is wide, high；
It does weighting by the friendship of each object candidate area and than three kinds of IOU, mutual information index and perceptual hash indexs and finds out often
The confidence level of one object candidate area, formula are as follows:
Confidence=1- (pHash+1/I (G(X), H(Y)/(IOU+D)
In formula, I (G(X), H(Y)) it is mutual information index, pHash is perceptual hash index, and for IOU to hand over and than index, D is setting
Constant, by the maximum object candidate area of confidence level alternately testing result.
4. the liquid level type Meter recognition method according to claim 3 based on crusing robot, which is characterized in that step
In 2.4, screening dial plate region method particularly includes: if the IOU of alternative testing result meets while being less than given threshold
ThresholdIOU, and (pHash+1/I (G(X), H(Y)) be greater than threshold value thresholdA when, by coarse positioning target instrument region make
For final goal instrument region, otherwise in case selecting testing result as final goal instrument region.
5. the liquid level type Meter recognition method according to claim 4 based on crusing robot, which is characterized in that step
In 2.4, threshold value thresholdIOU value range 0.1~0.4, threshold value thresholdA value range 10~50 are set.
6. the liquid level type Meter recognition method according to claim 1 based on crusing robot, which is characterized in that step 3
In, extract the specific steps in oil level region are as follows:
Step 3.1 carries out gray processing processing to dial plate image, uses histogram equalization, gaussian filtering removal interference；
Step 3.2 carries out opening operation, and deleting cannot disconnect narrow comprising the subject area of structural element, the profile of smooth object
Connection, remove tiny protrusion；
Step 3.3, using Otsu algorithm to dial plate region binaryzation；
Step 3.4 carries out subregion opening operation, first carries out opening operation to oil level pipe or so region, removes left and right noise；Again to upper
Lower region carries out opening operation, removes noise up and down；
Step 3.5, according to grey value difference, selected Threshold segmentation goes out oil level region, obtains oil level indicator region.
7. the liquid level type Meter recognition method according to claim 6 based on crusing robot, which is characterized in that step
In 3.5, setting threshold range is 40-70.
8. the liquid level type Meter recognition method according to claim 1 based on crusing robot, which is characterized in that step 4
In, calculate the specific steps of liquid level type instrument registration are as follows:
Step 4.1 carries out profile scan to oil level indicator region using Binarization methods, extracts largest connected domain；
Step 4.2 finds out dial plate intermediate region, and opening operation eliminates object block ambient noise, obtains object block position；
Step 4.3, according to the object block region divided, liquid level type instrument is calculated by oil level indicator region bound and object block position
Registration.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811150848.7A CN109359646A (en) | 2018-09-29 | 2018-09-29 | Liquid level type Meter recognition method based on crusing robot |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811150848.7A CN109359646A (en) | 2018-09-29 | 2018-09-29 | Liquid level type Meter recognition method based on crusing robot |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109359646A true CN109359646A (en) | 2019-02-19 |
Family
ID=65348208
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811150848.7A Pending CN109359646A (en) | 2018-09-29 | 2018-09-29 | Liquid level type Meter recognition method based on crusing robot |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109359646A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117237593A (en) * | 2023-10-08 | 2023-12-15 | 北京和利时系统工程有限公司 | Method and device for obtaining instrument value |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103927507A (en) * | 2013-01-12 | 2014-07-16 | 山东鲁能智能技术有限公司 | Improved multi-instrument reading identification method of transformer station inspection robot |
| CN105260412A (en) * | 2015-09-24 | 2016-01-20 | 东方网力科技股份有限公司 | Image storage method and device, and image retrieval method and device |
| CN106951930A (en) * | 2017-04-13 | 2017-07-14 | 杭州申昊科技股份有限公司 | A kind of instrument localization method suitable for Intelligent Mobile Robot |
| CN107610128A (en) * | 2017-09-26 | 2018-01-19 | 山东鲁能智能技术有限公司 | The method for inspecting and device of a kind of oil level indicator |
| CN107610162A (en) * | 2017-08-04 | 2018-01-19 | 浙江工业大学 | A kind of three-dimensional multimode state medical image autoegistration method based on mutual information and image segmentation |
-
2018
- 2018-09-29 CN CN201811150848.7A patent/CN109359646A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103927507A (en) * | 2013-01-12 | 2014-07-16 | 山东鲁能智能技术有限公司 | Improved multi-instrument reading identification method of transformer station inspection robot |
| CN105260412A (en) * | 2015-09-24 | 2016-01-20 | 东方网力科技股份有限公司 | Image storage method and device, and image retrieval method and device |
| CN106951930A (en) * | 2017-04-13 | 2017-07-14 | 杭州申昊科技股份有限公司 | A kind of instrument localization method suitable for Intelligent Mobile Robot |
| CN107610162A (en) * | 2017-08-04 | 2018-01-19 | 浙江工业大学 | A kind of three-dimensional multimode state medical image autoegistration method based on mutual information and image segmentation |
| CN107610128A (en) * | 2017-09-26 | 2018-01-19 | 山东鲁能智能技术有限公司 | The method for inspecting and device of a kind of oil level indicator |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117237593A (en) * | 2023-10-08 | 2023-12-15 | 北京和利时系统工程有限公司 | Method and device for obtaining instrument value |
| CN117237593B (en) * | 2023-10-08 | 2024-04-05 | 北京和利时系统工程有限公司 | Method and device for obtaining instrument value |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP5699788B2 (en) | Screen area detection method and system | |
| CN107543828B (en) | Workpiece surface defect detection method and system | |
| CN109801267B (en) | Inspection target defect detection method based on feature point detection and SVM classifier | |
| CN110148162A (en) | A kind of heterologous image matching method based on composition operators | |
| CN109447062A (en) | Pointer-type gauges recognition methods based on crusing robot | |
| CN109447061A (en) | Reactor oil level indicator recognition methods based on crusing robot | |
| CN108007388A (en) | A kind of turntable angle high precision online measuring method based on machine vision | |
| CN108256394A (en) | A kind of method for tracking target based on profile gradients | |
| CN110349207A (en) | A kind of vision positioning method under complex environment | |
| CN110807355A (en) | Pointer instrument detection and reading identification method based on mobile robot | |
| CN108520514B (en) | Consistency detection method for electronic elements of printed circuit board based on computer vision | |
| CN107240112B (en) | Individual X corner extraction method in complex scene | |
| CN112734761B (en) | Industrial product image boundary contour extraction method | |
| CN110263662B (en) | Human body contour key point and key part identification method based on grading | |
| CN109389165A (en) | Oil level gauge for transformer recognition methods based on crusing robot | |
| CN110047063B (en) | Material drop detection method, device, equipment and storage medium | |
| CN104778707A (en) | Electrolytic capacitor detecting method for improving general Hough transform | |
| CN110084830A (en) | A kind of detection of video frequency motion target and tracking | |
| CN104899888A (en) | Legemdre moment-based image subpixel edge detection method | |
| CN111563896A (en) | Image processing method for catenary anomaly detection | |
| CN110298344A (en) | A kind of positioning of instrument knob and detection method based on machine vision | |
| CN113705564B (en) | Pointer type instrument identification reading method | |
| CN107818583A (en) | Cross searching detection method and device | |
| CN110414521A (en) | Oil level gauge for transformer registration recognition methods in a kind of substation | |
| CN113673515A (en) | Computer vision target detection algorithm |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190219 |
|
| RJ01 | Rejection of invention patent application after publication |