CN104573713B - A kind of transformer Infrared image recognition based on image texture characteristic - Google Patents
A kind of transformer Infrared image recognition based on image texture characteristic Download PDFInfo
- Publication number
- CN104573713B CN104573713B CN201410850321.0A CN201410850321A CN104573713B CN 104573713 B CN104573713 B CN 104573713B CN 201410850321 A CN201410850321 A CN 201410850321A CN 104573713 B CN104573713 B CN 104573713B
- Authority
- CN
- China
- Prior art keywords
- image
- transformer
- point
- edge
- formula
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Landscapes
- Image Analysis (AREA)
Abstract
The present invention provides a kind of transformer Infrared image recognition based on image texture characteristic, gray processing and denoising are carried out to image first, and the Canny that X-direction is carried out to image is detected, after the step, the left and right edges textural characteristics of sleeve pipe are just exposed, by after the method for the matching of normalizated correlation coefficient maximum and K means clusters, left and right characteristic point is fitted respectively and is in line, determine that the highest marginal point on straight line and lowest edge point just determine the image-region of transformer, instant invention overcomes the shortcomings that traditional images recognition methods accuracy deficiency, anti-background interference ability is strong, the specific region of transformer in infrared thermal imagery can rapidly be identified, operations staff is helped to detect the position coordinates of transformer, improve the automaticity of POWER SYSTEM STATE monitoring, with very strong practical value.
Description
Technical field
The invention belongs to pattern-recognition and technical field of computer vision, more particularly to it is a kind of based on image texture characteristic
Transformer Infrared image recognition.
Background technology
Some Utilities Electric Co.s are mounted with video monitoring system in power plant, transformer station at present, and monitoring field apparatus, control can be achieved
The functions such as remote camera action processed.But these video monitoring systems only have video monitoring function, know automatically without video image
Other function.To give full play to the function of video monitoring system, more accurately judge the reason for accident alarm occurs for scene, should use
Remote digital video monitoring and digital image acquisition system, to realize the image recognition of equipment alarm, provided newly for accident detection
Means, provide reliable foundation for crash analysis.
At present both at home and abroad to the still planless discussion of transformer automatic identification based on infrared image textural characteristics.Tang Hui
It is bright, Zhang Jian《Remote image supervision of sub station system design》Article proposes a kind of electrical equipment based on image Hu not bending moments
Image-recognizing method, this method calculate the Hu of the bianry image not bending moment, constitute the characteristic vector of power equipment, devise
BP neural network grader does Classification and Identification, available for power equipment infrared image in combination temperature information realization power system
Fault Identification.But this method, which only can recognize that in image, has certain type of electrical equipment, and fails to realize electrical equipment
It is accurately positioned, therefore has certain limitation.Chen Junyou, Gionee army, Duan Shaohui, Yao Senjing, Zhao Ling survey article《Based on Hu not bending moments
Infrared image power equipment identification》It is proposed that one kind utilizes infrared image and neural network weight Weigh Direct Determination to carry out scene zero
It is worth insulator and knows method for distinguishing, but it is only limitted to the identification field of insulator, is not directed to transformer identification, while this paper side
Method is only applicable to the few infrared image of ambient interferences, and the actual conditions of this and substation field are not inconsistent, live bad adaptability.This
Outside, most of achievements in research are based primarily upon the gray difference between transformer and background to be extracted and be identified to target.But
Due to ambient interferences, the degree of accuracy of identification is not high.For this reason, it is necessary to develop a kind of for the infrared of transformer texture feature extraction
Image-recognizing method to improve the accuracy of identification and specific aim, but existing document or pay close attention to the overall identification classification of image and
Being accurately positioned for target, or the positioning of concern target device can not be realized, the research of accuracy is identified to improving but more to be had
Limit.
Image recognition based on image texture characteristic extraction has been applied in pattern-recognition multiple fields, is wrapped
Include machine components identification, Aero-Space, agricultural engineering etc..Image-recognizing method for transformer itself texture feature extraction is
In infrared image transformer be accurately positioned and identification provides the foundation.
The content of the invention
The present invention proposes a kind of transformer Infrared image recognition based on image texture characteristic, it is therefore intended that improves
Traditional images recognition methods accuracy.
It is contemplated that overcome the deficiencies in the prior art, the characteristics of taking into full account substation field background complexity, fully examine
Consider the image texture characteristic of transformer edge " waveform ", incorporated the matching of normalizated correlation coefficient maximum and K-means gathers
Class method, proposes a kind of transformer Infrared image recognition based on image texture characteristic, and this method recognition result being capable of visitor
The position coordinate data of ground reaction transformer in the picture is seen, is laid the foundation for follow-up temperature and Fault Identification, lifts power network
The automaticity of fault distinguishing, to reach above-mentioned purpose, the technical solution adopted by the present invention is：One kind is special based on image texture
The transformer Infrared image recognition of sign, comprises the following steps：
(1) image preprocessing, including carry out X-direction to transformer image gray processing and denoising, and to image
Canny is detected, and extracts the feature texture of transformer；
(2) pretreatment image and left and right edges characteristic feature image are entered using the method for normalizated correlation coefficient maximum
Row matching, realize the texture characteristic points extraction of image；
(3) K-means clustering processings are carried out to characteristic point, and rejects outlier；
(4) according to the cluster result of left and right edges, the fitting a straight line of least square method is carried out to cluster centre, determines sleeve pipe
Left and right edges；
(5) ordinate highest and minimum point are found on the edge line of fitting, determines the lower edges of sleeve pipe, it is final to obtain
To the transformer region of identification.
Further, the step (1) specifically comprises the following steps：
A) method of image gray processing processing is as follows：
Gray (i, j)=0.299 × r (i, j)+0.587 × g (i, j)+0.114 × b (i, j) formula (1)
Wherein, gray (i, j) represents the gray value of (i, j) point, and r (i, j) represents the R value components of (i, j) point rgb value, g
(i, j) represents the G value components of (i, j) point rgb value, and b (i, j) represents the B value components of (i, j) point rgb value；
B) method that image denoising uses Butterworth low pass ripple, the transmission function of a n rank Butterworth filter
For：
Wherein, D0It is off frequency；N is exponent number, takes positive integer, so as to the shape of controlling curve, because Butterworth is filtered
Ripple device transfer curve is more smooth, therefore the fuzzy of image will be reduced；
C) the x direction gradient values of image are after pretreatment：
grayx(i, j)=gray (i+1, j)-gray (i-1, j) formula (3)
In formula (3), the center pixel of neighborhood is compared with the interpolation result of two gradient magnitudes along gradient direction,
If the amplitude of the centre of neighbourhood is big unlike two interpolation results of gradient direction, by edge flag position corresponding to gray (i, j)
It is entered as 0.
Further, the step (2) specifically comprises the following steps：
A) normalizated correlation coefficient is calculated, computational methods are as follows：
Wherein
In formula, f (x, y) is M × N original image, i.e., by the X-direction Canny detection images of pretreatment, t (p, q) is P
× Q (P≤M, Q≤N) template image, RN (x, y) are normalizated correlation coefficient；
B) RN (x, y) ＜ 0.7 point is filtered out, remaining point is considered as the left and right edges feature of transformer.
Further, the step (3) specifically comprises the following steps：
A) method of K-means clusters is as follows：
K' is clusters number, selectes 14 cluster centres, the initial cluster center of 14 classes is randomly choosed first, to any
One object, the distance to this 14 centers is calculated, and the object is grouped into the class where the maximum center of similitude, utilized
Averaging method updates the central value of k class, to all 14 cluster centres, if after iterative algorithm updates, shown in formula (6)
Measure function convergence or reach maximum iteration, then iteration terminates, and otherwise continues iteration；
B) coordinate value of 14 cluster centres after K-means is clustered is set as (iw,jw), wherein w=1,2...14,
Respectively calculate 1~14 cluster centre point arrive other 13 cluster centre points Euclidean distance, and formation 14 × 14 ranks it is European away from
From matrix, Euclidean distance matrix by rows is added to form Euclidean distance and matrix dis (u, 1) u=1,2...14, then by Euclidean away from
From with arrangement form dis'(u from small to large, 1) u=1,2...14 matrixes, whenWhen, assert
U points are outlier.
Further, the step (4) specifically comprises the following steps：
A) using the cluster centre point set coordinate for rejecting outlier as sample, fitting a straight line, tool are carried out using least square method
The formula of body is as follows：
Wherein, o represents that 14 cluster centre points reject the cluster centre number after outlier, xi,yiRespectively cluster centre
The transverse and longitudinal coordinate value of point,The respectively average value of cluster centre point transverse and longitudinal coordinate, (9) formula are the linear equation of fitting；
B) the transformer left and right edges characteristic point for adding RN (x, y) ＜ 0.7 carrys out the degree of accuracy of accentuated edges identification, carries out side
Edge is revised：It is assumed that φ { iiss,jjssIt is edge feature point set, ss is the numbering of Edge Feature Points, iiss,jjssRespectively its
Transverse and longitudinal coordinate value, if
Think (iiss,jjss) it is actual Edge Feature Points, (10) formula arrives a range formula for straight line for point, by reality
Checking is real, when selecting Edge Feature Points distance feature air line distance as within 3 pixels, can include most Edge Feature Points,
Cause minimal error simultaneously；If do not meet (10) formula, then it is assumed that it is the Edge Feature Points to peel off, it is final to form actual edge feature
Point set Φ { iiss,jjss, bring the coordinate at the set midpoint into (7)~(9) formula, finally give the edge line side of amendment
Journey.
Further, the step (5) specifically comprises the following steps：
The highs and lows ordinate of transformer left and right edges actual edge characteristic point is determined, so that it is determined that sleeve pipe
The straight line of lower edges, do two straight lines of lower edges and be fitted amendment straight line intersection with left and right edges, middle region is
The transformer region finally identified.
The present invention has the advantages and positive effects of：It is red by the transformer proposed by the present invention based on image texture characteristic
Outer image-recognizing method, the feature texture of transformer can be extracted according to " waveform " feature of transformer left and right edges,
After K-means clusters and outlier are rejected, left and right edges straight line is fitted to, and more multiple edge letter is contained according to feature texture
The characteristics of breath, corrects edge line, and the present invention can rapidly identify the specific region of transformer in infrared thermal imagery, help to run
Personnel detect the position coordinates of transformer, improve the automaticity of POWER SYSTEM STATE monitoring, have very strong practical value.
Brief description of the drawings
Fig. 1 is the schematic diagram of the transformer infrared image of substation field shooting；
Fig. 2 is the schematic diagram of transformer infrared image image after pretreatment；
Fig. 3 is the schematic diagram of the template image of transformer left and right edges feature；
Fig. 4 is the schematic diagram of the transformer left and right edges feature dot image after the matching of normalizated correlation coefficient maximum；
Fig. 5 is the schematic diagram of result schematic diagram of the transformer left hand edge characteristic point after K-means is clustered；
Fig. 6 is the schematic diagram of result schematic diagram of the transformer right hand edge characteristic point after K-means is clustered；
Fig. 7 is the schematic diagram of the Euclidean distance matrix schematic diagram between cluster centre point；
Fig. 8 is the schematic diagram of the schematic diagram of left and right edges fitting a straight line；
Fig. 9 is the schematic diagram for adding after Edge Feature Points the fitted straight lines of edges schematic diagram corrected；
Figure 10 is the schematic diagram of position view of the transformer in infrared image.
Embodiment
It is applied to Tianjin below by way of by the transformer Infrared image recognition based on image texture characteristic of the present invention
In the transformer infrared image identification of Baodi Utilities Electric Co. collection.
In the identifying system, including Baodi power supply administration has transformer station's transformer equipment under its command, covers vegetable garden great Kou Tun, city
Guan Xin openings, the old mouth Zhou Liang in east, with village, peaceful village Deng Di transformer stations transformer equipment, be taken on site altogether including 320 mutual
Sensor infrared image, the transformer size and angle in these infrared images are variant.The present invention proposes transformer identification side
Method, for identifying the particular location of transformer in these infrared images, and extract from image so as to follow-up temperature and
Fault Identification, comprise the following steps that：
Step 1：Transformer infrared image in figure (1) is pre-processed, removes digital information and the right side of lower section first
Side detects, carried than vitta region, then progress image gray processing and denoising, the Canny that X-direction is finally carried out to image
The feature texture of transformer is taken out, completes the pretreatment of image.
The pretreatment of image specifically comprises the following steps：
A) method of image gray processing processing is as follows：
Gray (i, j)=0.299 × r (i, j)+0.587 × g (i, j)+0.114 × b (i, j) formula (1)
Wherein, gray (i, j) represents the gray value of (i, j) point, and r (i, j) represents the R value components of (i, j) point rgb value, g
(i, j) represents the G value components of (i, j) point rgb value, and b (i, j) represents the B value components of (i, j) point rgb value.
B) method that image denoising uses Butterworth low pass ripple, the transmission function of a n rank Butterworth filter
For：
Wherein, D0It is off frequency；N is exponent number, takes positive integer, so as to the shape of controlling curve, because Butterworth is filtered
Ripple device transfer curve is more smooth, therefore the fuzzy of image will be reduced.
C) the x direction gradient values of image are after pretreatment：
grayx(i, j)=gray (i+1, j)-gray (i-1, j) formula (3)
In formula (3), the center pixel of neighborhood is compared with the interpolation result of two gradient magnitudes along gradient direction,
If the amplitude of the centre of neighbourhood is big unlike two interpolation results of gradient direction, by edge flag position corresponding to gray (i, j)
It is entered as 0.
Final result is as schemed shown in (2).
Step 2：Using the method for normalizated correlation coefficient maximum to pretreatment image and left and right edges characteristic feature figure
As being matched, the texture characteristic points extraction of image is realized.Figure (3) is the transformer left and right edges feature templates image of standard,
To scheme (2) for original image,
If f (x, y) is M × N original image, i.e., by the X-direction Canny detection images of pretreatment, t (p, q) is P × Q
The template image of (P≤M, Q≤N), then coefficient R N (x, y) can be expressed as：
Wherein
After being calculated through formula (4)~(5), the point that normalizated correlation coefficient is less than 0.7 is removed, as a result as shown in figure (4).
Step 3：K-means clusters are carried out respectively to all left hand edge characteristic points and right hand edge characteristic point, gathered respectively for 14
Class, forms 28 cluster centre points altogether, and method is as follows：
Cluster result is as schemed shown in (5)~(6).Then Euclidean distance matrix is calculated, if 14 after K-means is clustered
The coordinate value of individual cluster centre is (iw,jw), wherein w=1,2...14, calculate respectively 1~14 cluster centre point to other 13
The Euclidean distance of individual cluster centre point, and 14 × 14 rank Euclidean distance matrixes are formed, Euclidean distance matrix by rows is added to be formed
Euclidean distance and matrix dis (u, 1) u=1,2...14, then by Euclidean distance and arrangement form dis'(u from small to large, 1) u=
1,2...14 matrix, whenWhen, assert that u points are outlier.As a result as shown in figure (7), this is mutual
There are three outliers in sensor left hand edge characteristic point, after outlier is rejected, form final cluster centre set.
Step 4：Least square method is carried out to left hand edge cluster centre meeting point and right hand edge cluster centre meeting point respectively
Fitting a straight line, specific formula are as follows：
Wherein, o represents that 14 cluster centre points reject the cluster centre number after outlier, xi,yiRespectively cluster centre
The transverse and longitudinal coordinate value of point,The respectively average value of cluster centre point transverse and longitudinal coordinate, (9) formula are the linear equation of fitting, as a result
As shown in figure (8), to improve the degree of accuracy of limb recognition, the transformer left and right edges characteristic point for adding RN (x, y) ＜ 0.7 is come by force
Change the degree of accuracy of limb recognition, and because the straight line of fitting is approximate accurate, therefore edge amendment can be carried out by it based on, it is assumed that
φ{iiss,jjssIt is edge feature point set, ss is the numbering of Edge Feature Points, iiss,jjssRespectively its transverse and longitudinal coordinate value,
If
Think (iiss,jjss) it is actual Edge Feature Points, if not meeting (10) formula, then it is assumed that be the edge to peel off
Characteristic point, it is final to form actual edge set of characteristic points Φ { iiss,jjss, bring the coordinate at the set midpoint into (7)~(9)
Formula, the edge line equation of amendment is finally given, fitting a straight line result is modified by formula (10), as a result as shown in figure (9).
Step 5：It is assumed that the highs and lows ordinate of transformer left hand edge actual edge characteristic point divides leftmax,
leftmin, the ordinate of the highs and lows of right hand edge actual edge characteristic point is respectively rightmax,rightmin。
To sum up, top edge peak ordinate max=max { leftmax,rightmax, lower edge minimum point ordinate min
=min { leftmin,rightmin, two straight lines of y=max and y=min are done respectively and are fitted amendment straight line phase with left and right edges
Hand over, the transformer region that middle region as finally identifies, transformer position view is as schemed shown in (10).
Embodiments of the invention are described in detail above, but the content is only presently preferred embodiments of the present invention,
It is not to be regarded as the practical range for limiting the present invention.All equivalent changes made according to the scope of the invention and improvement etc., all should
Still belong within this patent covering scope.
Claims (6)
1. a kind of transformer Infrared image recognition based on image texture characteristic, comprises the following steps：
(1) image preprocessing, including to transformer image gray processing and denoising, and the Canny that X-direction is carried out to image is examined
Survey, extract the feature texture of transformer；
(2) using the method for normalizated correlation coefficient maximum to pretreatment image and the progress of left and right edges characteristic feature image
Match somebody with somebody, realize the texture characteristic points extraction of image；
(3) K-means clustering processings are carried out to characteristic point, and rejects outlier；
(4) according to the cluster result of left and right edges, the fitting a straight line of least square method is carried out to cluster centre, determines a left side for sleeve pipe
Right hand edge；
(5) ordinate highest and minimum point are found on the edge line of fitting, the lower edges of sleeve pipe is determined, finally gives knowledge
Other transformer region.
2. a kind of transformer Infrared image recognition based on image texture characteristic according to claim 1, its feature
It is：The step (1) specifically comprises the following steps：
A) method of image gray processing processing is as follows：
Gray (i, j)=0.299 × r (i, j)+0.587 × g (i, j)+0.114 × b (i, j) formula (1)
Wherein, gray (i, j) represents the gray value of (i, j) point, and r (i, j) represents the R value components of (i, j) point rgb value, g (i, j)
The G value components of (i, j) point rgb value are represented, b (i, j) represents the B value components of (i, j) point rgb value；
B) method that image denoising uses Butterworth low pass ripple, the transmission function of a n rank Butterworth filter are：
Wherein, D0It is off frequency；N is exponent number, takes positive integer, so as to the shape of controlling curve, because Butterworth filter turns
Shifting curve is more smooth, therefore the fuzzy of image will be reduced；
C) the x direction gradient values of image are after pretreatment：
grayx(i, j)=gray (i+1, j)-gray (i-1, j) formula (3)
In formula (3), the center pixel of neighborhood is compared with the interpolation result of two gradient magnitudes along gradient direction, such as
The amplitude of the fruit centre of neighbourhood is big unlike two interpolation results of gradient direction, then assigns edge flag position corresponding to gray (i, j)
It is worth for 0.
3. a kind of transformer Infrared image recognition based on image texture characteristic according to claim 1, its feature
It is：The step (2) specifically comprises the following steps：
A) normalizated correlation coefficient is calculated, computational methods are as follows：
Wherein
In formula, f (x, y) is M × N original image, i.e., by the X-direction Canny detection images of pretreatment, t (p, q) is P × Q (P
≤ M, Q≤N) template image, RN (x, y) is normalizated correlation coefficient；
B) RN (x, y) ＜ 0.7 point is filtered out, remaining point is considered as the left and right edges feature of transformer.
4. a kind of transformer Infrared image recognition based on image texture characteristic according to claim 1, its feature
It is：The step (3) specifically comprises the following steps：
A) method of K-means clusters is as follows：
K' is clusters number, selectes 14 cluster centres, the initial cluster center of 14 classes is randomly choosed first, to any one
Object, the distance to this 14 centers is calculated, and the object is grouped into the class where the maximum center of similitude, utilize average
Method updates the central value of k class, to all 14 cluster centres, if after iterative algorithm updates, the survey shown in formula (6)
Degree function convergence reaches maximum iteration, then iteration terminates, and otherwise continues iteration；
B) coordinate value of 14 cluster centres after K-means is clustered is set as (iw,jw), wherein w=1,2...14, respectively
1~14 cluster centre point is calculated to the Euclidean distance of other 13 cluster centre points, and forms 14 × 14 rank Euclidean distance squares
Battle array, Euclidean distance matrix by rows is added to form Euclidean distance and matrix dis (u, 1), u=1,2...14, then by Euclidean distance
Arrangement form dis'(u from small to large, 1), u=1,2...14, matrix, whenWhen, assert
U points are outlier.
5. a kind of transformer Infrared image recognition based on image texture characteristic according to claim 1, its feature
It is：The step (4) specifically comprises the following steps：
A) using the cluster centre point set coordinate for rejecting outlier as sample, fitting a straight line is carried out using least square method, specifically
Formula is as follows：
Wherein, o represents that 14 cluster centre points reject the cluster centre number after outlier, xi,yiRespectively cluster centre point
Transverse and longitudinal coordinate value,The respectively average value of cluster centre point transverse and longitudinal coordinate, (9) formula are the linear equation of fitting；
B) the transformer left and right edges characteristic point for adding RN (x, y) ＜ 0.7 carrys out the degree of accuracy of accentuated edges identification, carries out edge and repaiies
Order：It is assumed that φ { iiss,jjssIt is edge feature point set, ss is the numbering of Edge Feature Points, iiss,jjssRespectively its transverse and longitudinal
Coordinate value, if
Think (iiss,jjss) it is actual Edge Feature Points, (10) formula is a range formula for point to straight line, is demonstrate,proved by testing
It is real, when selecting Edge Feature Points distance feature air line distance as within 3 pixels, most Edge Feature Points can be included, simultaneously
Cause minimal error；If do not meet (10) formula, then it is assumed that it is the Edge Feature Points to peel off, it is final to form actual edge characteristic point
Set Φ { iiss,jjss, bring the coordinate at the set midpoint into (7)~(9) formula, finally give the edge line equation of amendment.
6. a kind of transformer Infrared image recognition based on image texture characteristic according to claim 1, its feature
It is：The step (5) specifically comprises the following steps：
The highs and lows ordinate of transformer left and right edges actual edge characteristic point is determined, so that it is determined that sleeve pipe is upper and lower
The straight line at edge, do two straight lines of lower edges and be fitted amendment straight line intersection with left and right edges, middle region is as final
The transformer region of identification.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410850321.0A CN104573713B (en) | 2014-12-31 | 2014-12-31 | A kind of transformer Infrared image recognition based on image texture characteristic |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410850321.0A CN104573713B (en) | 2014-12-31 | 2014-12-31 | A kind of transformer Infrared image recognition based on image texture characteristic |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104573713A CN104573713A (en) | 2015-04-29 |
| CN104573713B true CN104573713B (en) | 2017-12-26 |
Family
ID=53089733
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410850321.0A Active CN104573713B (en) | 2014-12-31 | 2014-12-31 | A kind of transformer Infrared image recognition based on image texture characteristic |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104573713B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109215364A (en) * | 2018-11-19 | 2019-01-15 | 长沙智能驾驶研究院有限公司 | Traffic signals recognition methods, system, equipment and storage medium |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106327499B (en) * | 2016-08-30 | 2017-09-26 | 南京鑫和汇通电子科技有限公司 | The identification of greasy dirt image based on marginal point self-similarity and TEDS systems |
| CN106682079B (en) * | 2016-11-21 | 2020-06-05 | 云南电网有限责任公司电力科学研究院 | User electricity consumption behavior detection method based on cluster analysis |
| CN108955901B (en) * | 2017-07-06 | 2020-03-31 | 深圳市朗驰欣创科技股份有限公司 | Infrared temperature measurement method and system and terminal equipment |
| CN107944453A (en) * | 2017-12-05 | 2018-04-20 | 江苏省电力试验研究院有限公司 | Based on Hu not bushing detection methods of bending moment and support vector machines |
| CN109040598B (en) * | 2018-08-29 | 2020-08-14 | Oppo广东移动通信有限公司 | Image processing method, image processing device, computer-readable storage medium and electronic equipment |
| CN110321934B (en) * | 2019-06-12 | 2023-05-12 | 深圳供电局有限公司 | Method and system for detecting abnormal data of user electricity consumption |
| CN111832571B (en) * | 2020-07-09 | 2021-03-05 | 哈尔滨市科佳通用机电股份有限公司 | Automatic detection method for truck brake beam strut fault |
| CN115691026B (en) * | 2022-12-29 | 2023-05-05 | 湖北省林业科学研究院 | Intelligent early warning monitoring management method for forest fire prevention |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103837135A (en) * | 2012-11-27 | 2014-06-04 | 国际商业机器公司 | Workpiece detecting method and system |
-
2014
- 2014-12-31 CN CN201410850321.0A patent/CN104573713B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103837135A (en) * | 2012-11-27 | 2014-06-04 | 国际商业机器公司 | Workpiece detecting method and system |
Non-Patent Citations (2)
| Title |
|---|
| 几种边缘检测算法在变压器故障红外诊断图像处理中的应用;徐步洋等;《机电工程技术》;20081031;第37卷(第10期);第108-110页 * |
| 基于聚类划分的两阶段离群点检测算法;杨福萍等;《计算机应用研究》;20130731;第30卷(第7期);第1942-1945页 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109215364A (en) * | 2018-11-19 | 2019-01-15 | 长沙智能驾驶研究院有限公司 | Traffic signals recognition methods, system, equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104573713A (en) | 2015-04-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104573713B (en) | A kind of transformer Infrared image recognition based on image texture characteristic | |
| CN101620676B (en) | Fast image recognition method of insulator contour | |
| CN107392963B (en) | Eagle eye-imitated moving target positioning method for soft autonomous aerial refueling | |
| CN105023014B (en) | A kind of shaft tower target extraction method in unmanned plane inspection transmission line of electricity image | |
| CN104361314B (en) | Based on infrared and transformer localization method and device of visual image fusion | |
| CN108564565A (en) | A kind of power equipment infrared image multi-target orientation method based on deep learning | |
| CN106339707B (en) | A kind of gauge pointer image-recognizing method based on symmetric characteristics | |
| CN111536981B (en) | Embedded binocular non-cooperative target relative pose measurement method | |
| CN106296700B (en) | A kind of steel cord conveyor belt connector twitch detection method | |
| CN106056594A (en) | Double-spectrum-based visible light image extraction system and method | |
| CN107492094A (en) | A kind of unmanned plane visible detection method of high voltage line insulator | |
| CN109447949A (en) | Insulated terminal defect identification method based on crusing robot | |
| CN109615598A (en) | A kind of power transmission line recognition methods based on the free algorithm of edge mapping parameter | |
| CN103903238A (en) | Method for fusing significant structure and relevant structure of characteristics of image | |
| CN107481237A (en) | A kind of infrared array image hot spot detection method based on multiframe temperature characterisitic | |
| CN110245592A (en) | A method of for promoting pedestrian's weight discrimination of monitoring scene | |
| CN112132157A (en) | Raspberry pie-based gait face fusion recognition method | |
| CN110320523A (en) | Follow the target locating set and method of robot | |
| CN101561316B (en) | On-line test visual data processing system based on region of interest (ROI) | |
| CN111985436A (en) | Workshop ground mark line identification fitting method based on LSD | |
| CN106919929A (en) | Insulator automatic identifying method in a kind of infrared image based on template matches | |
| CN111340884A (en) | Binocular heterogeneous camera and RFID dual target positioning and identity identification method | |
| CN113160236B (en) | Image identification method for shadow shielding of photovoltaic cell | |
| Sun et al. | Identification and localization method of the insulator based on three-dimensional point cloud modeling | |
| CN110956640B (en) | Heterogeneous image edge point detection and registration method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant |