CN106156771A - A kind of meter reading Region detection algorithms based on multi-feature fusion - Google Patents
A kind of meter reading Region detection algorithms based on multi-feature fusion Download PDFInfo
- Publication number
- CN106156771A CN106156771A CN201610513983.8A CN201610513983A CN106156771A CN 106156771 A CN106156771 A CN 106156771A CN 201610513983 A CN201610513983 A CN 201610513983A CN 106156771 A CN106156771 A CN 106156771A
- Authority
- CN
- China
- Prior art keywords
- image
- feature
- water meter
- characteristic
- detection algorithms
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
- G06V20/63—Scene text, e.g. street names
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
One disclosed by the invention meter reading based on multi-feature fusion Region detection algorithms, comprises the following steps: S1, acquisition training data;S2, cut out reading area and non-reading area in water meter image, extract the multi-channel feature of this cutting zone and carry out Feature Fusion, with this feature for input training objective grader;S3, extraction water meter image multi-channel feature, calculate its characteristic-integration figure;S4, utilize characteristic-integration figure calculate each sliding window feature, with fusion feature for input, utilize S2 training gained grader sliding window is classified, obtain target window;S5, with extrapolation method estimate artwork characteristic pattern under multiple yardsticks, repeat S4, S5, obtain multiple dimensioned target window;S6, rotation artwork, repeat S3, S4, S5, S6, obtain multidirectional target window.The invention provides a kind of accurate, robust, practical meter reading Region detection algorithms.
Description
Technical field
The present invention relates to artificial intelligence field, calculate particularly to a kind of meter reading region detection based on multi-feature fusion
Method.
Background technology
In recent years, along with the development of mobile Internet and popularizing of digital product, come from distinct device (intelligence hands
The photographic head such as machine, digital camera, even automatic Pilot streetscape car, unmanned plane) view data continue explosion type ground increase.These
In the image of magnanimity, there is quite a few view data to carry Word message, and Word message generally contains and is highly profitable
Semantic information.Such as, these Word messages are probably the description to building, shop, traffic sign, guideboard, trade name etc..
Therefore, the semantic information of these high levels can be widely applied to machine reading, automatic bat is translated, image retrieval, video frequency searching, language
The occasions such as speech translation, automatic Pilot, robot navigation.The mankind need the vision character analysis skill of a kind of intelligence more urgently
Art.Vision character analysis is a kind of technology extracting and understanding Word message from the angle of machine vision.It relates at image
A series of subject knowledges such as reason, pattern recognition, computer vision, machine learning and psychology, are the most all association areas
One of important research direction.
Water meter automatic reading based on computer vision is exactly an important application in vision character analysis, and it can take
For existing artificial meter reading mode so that meter reading becomes automatic flow.The text analyzing of view-based access control model primarily solves
Problem certainly is exactly the detection of character area.
The problem that the water meter automatic reading of view-based access control model primarily solves is exactly the detection of reading area, the method for current main flow
It is method based on image procossing, is examined by image denoising, image binaryzation based on color characteristic, direction based on line detection
The steps such as survey, region segmentation determine reading area.But this method is to the illumination under various complex scenes, deformation, coverage
Bad etc. condition adaptability, easily it is disturbed, poor robustness.
Summary of the invention
It is an object of the invention to the shortcoming overcoming prior art with not enough, it is provided that a kind of water meter based on multi-feature fusion
Reading area detection algorithm.
The purpose of the present invention is realized by following technical scheme:
A kind of meter reading Region detection algorithms based on multi-feature fusion, comprises the steps of
S1, acquisition training data, shoot water meter image pattern by photographic head, and enter the meter reading region in image
Pedestrian's work marks, and obtains the center of meter reading, length and width information;
Reading area and non-reading area in S2, cutting water meter image, the multi-channel feature extracting this cutting zone is gone forward side by side
Row Feature Fusion, with extract multi-channel feature for input training image grader;Described cutting zone is uncertain region,
Classified by Image Classifier, be divided into reading area and non-reading area;
S3, extract water meter image multi-channel feature, described multi-channel feature include gradient orientation histogram, gradient magnitude,
LUV color characteristic, greyscale color feature, calculate the characteristic-integration figure of water meter image;
S4, traveling through whole sliding window, utilize characteristic-integration figure to calculate each sliding window feature, the image utilizing S2 train is classified
Sliding window is classified by device, obtains target window;
S5, with extrapolation method estimate artwork characteristic-integration figure under multiple yardsticks, repeat S4, S5 step, obtain many chis
The target window of degree;
S6, rotation artwork, repeat S3, S4, S5, S6 step, obtain multidirectional target window.
Described step S1 particularly as follows:
S1.1, by the water meter image pattern in RGB camera collection actual scene;
S1.2, the meter reading region in water meter image pattern acquired in S1.1 is carried out artificial mark, including
The center in meter reading region (x, y), length h, width w and angle a.
Described water meter image pattern includes following different parameter: illumination, visual angle, water meter type, the water meter extent of damage.This
Sample is to ensure that the multiformity of sample.
Described step S2 particularly as follows:
S2.1, mark according to step S1 gained water meter image and reading area, cut out reading area and non-reading, be used for
The classification of target;
S2.2, the multi-channel feature of extraction S2.1 institute cutting image, with multi-channel feature for input, train integrated decision tree
Whether grader is to classify in reading area to this region.
Described step S3 particularly as follows:
S3.1, extract water meter image multi-channel feature, including gradient orientation histogram, gradient magnitude, LUV color characteristic,
Greyscale color feature, and calculate characteristic-integration figure;
The computational methods of described gradient direction:
Wherein (i j) is pixel (i, j) angle of place's gradient direction to O;I is image, and x represents horizontal direction, and y represents and hangs down
Nogata to, i represents pixel coordinate in the horizontal direction;J represents pixel coordinate in vertical direction;
The computational methods of described gradient magnitude:
Wherein (i j) is the gradient magnitude at pixel (i.j) place to M;I is image, and x represents horizontal direction, and y represents Vertical Square
To, i represents pixel coordinate in the horizontal direction;J represents pixel coordinate in vertical direction.
Described step S4 particularly as follows:
S4.1, travel through each sliding window, utilize step S3 gained characteristic-integration figure, calculate the multi-channel feature of each sliding window,
And carry out Feature Fusion;
S4.2, with S4.1 merge characteristic vector for input, utilize S2 training gained Image Classifier sliding window is carried out
Classification, obtains sliding window significance;
S4.3, the target detected is carried out maximization suppression, obtain detect target.
Described step S5 particularly as follows:
S5.1, utilize the statistical property of adjacent scalogram picture, by the method for extrapolation estimate multiple dimensioned under artwork feature,
Extrapolation algorithm is as follows:
The wherein yardstick of S representative image, CsRepresenting the characteristics of image under yardstick S, (C is s) that characteristic pattern C is carried out yardstick to R
Resampling for S;λ is the channel parameters that the statistical analysis by great amount of images obtains;Ω represents different image channels;Extrapolation
Algorithm quickly calculates the characteristics of image of multiple adjacent yardstick by the characteristics of image calculated under a certain fixed size;
S5.2, repeat S4, S5, obtain multiple dimensioned under reading area target window.
Described step S6 particularly as follows:
With 5 degree for interval, multiple rotary original water gauge image, repeats S3, S4, S5, S6 step, is calculated multidirectional water
Meter reading region, i.e. obtains multidirectional target window.
The present invention compared with prior art, has the advantage that and beneficial effect:
1, the present invention utilizes the Multi resolution feature extraction method of multiple features fusion, enriching under the multiple yardsticks of efficient extraction
Information, solves the primary difficult problem in water meter automatic reading, meter reading region detection the most multiple dimensioned, multidirectional, has weight
Big application prospect.
2, the present invention uses the thinking of target detection, utilizes Multi resolution feature extraction method, the multiple yardsticks of efficient extraction
Under abundant information, thus under multi-angle, multiple dimensioned image, carry out sliding window scanning rapidly, obtain confidence level maximum
Reading area target.It addition, the method that present invention employs multiple features fusion, therefore algorithm robustness is strong, to various complicated fields
The adaptability of scape is good.The present invention solves the reading area test problems of the multi-direction water meter under complex scene, for further
Recognition of Reading provides the foundation.
Accompanying drawing explanation
Fig. 1 is the flow chart of a kind of meter reading Region detection algorithms based on multi-feature fusion of the present invention.
Fig. 2 is the training flow chart of algorithm described in Fig. 1.
Detailed description of the invention
Below in conjunction with embodiment and accompanying drawing, the present invention is described in further detail, but embodiments of the present invention do not limit
In this.
As it is shown in figure 1, meter reading Region detection algorithms based on multi-feature fusion mainly comprises the steps:
S1, acquisition training data;Training process is as shown in Figure 2;
S1.1, by the water meter image pattern in a large amount of actual scene of RGB camera collection, including various light
According to, visual angle, water meter type, water meter extent of damage etc., to ensure the multiformity of sample;
S1.2, the meter reading region in water meter image pattern acquired in S1.1 is carried out artificial mark, including
The center in meter reading region (x, y), length (h), width (w) and angle (a);
S2 training objective grader:
S2.1, mark according to S1 gained water meter image and reading area, cut out reading area and non-reading, for target
Classification;
S2.2, the multi-channel feature of extraction S2.1 institute cutting image, with this feature for input, train integrated decision tree classification
Whether device is to classify in reading area to this region;
S3 calculates image multi-channel feature:
S3.1, extract water meter image multi-channel feature, including gradient orientation histogram, gradient magnitude, LUV color characteristic,
Greyscale color feature, and calculate characteristic-integration figure:
Gradient direction computational methods:
Gradient magnitude computational methods:
S4 sliding window scans:
S4.1, travel through each sliding window, utilize S3.1 gained characteristic-integration figure, calculate the multi-channel feature of each sliding window, and
Carry out Feature Fusion;
S4.2, with S4.1 merge characteristic vector for input, utilize S2 training gained grader sliding window is classified,
Obtain sliding window significance;
S4.3, the target detected is carried out maximization suppression, obtain detect target.
S5 extrapolation Analysis On Multi-scale Features:
S5.1, utilize the statistical property of adjacent scalogram picture, by the method for extrapolation estimate multiple dimensioned under artwork feature,
Extrapolation algorithm is as follows:
The wherein yardstick of S representative image, CsRepresenting the characteristics of image under yardstick S, (C is s) that characteristic pattern C is carried out yardstick to R
Resampling for S;λ is the channel parameters that the statistical analysis by image obtains;Ω represents different image channels;Extrapolation algorithm
The characteristics of image of multiple adjacent yardstick is quickly calculated by the characteristics of image calculated under a certain fixed size;
S5.2, repeat S4, S5, obtain multiple dimensioned under reading area target window.
S6 rotates detection:
S6.1, with 5 degree for interval, multiple rotary original water gauge image, repeat S3, S4, S5, S6 step, can be calculated
Multidirectional meter reading region.
Above-described embodiment is the present invention preferably embodiment, but embodiments of the present invention are not by above-described embodiment
Limit, the change made under other any spirit without departing from the present invention and principle, modify, substitute, combine, simplify,
All should be the substitute mode of equivalence, within being included in protection scope of the present invention.
Claims (8)
1. a meter reading Region detection algorithms based on multi-feature fusion, it is characterised in that comprise the steps of
S1, acquisition training data, shoot water meter image pattern by photographic head, and the meter reading region in image carried out people
Work marks, and obtains the center of meter reading, length and width information;
Reading area and non-reading area in S2, cutting water meter image, extract the multi-channel feature of this cutting zone and carry out spy
Levy fusion, with extract multi-channel feature for input training image grader;Described cutting zone is uncertain region, by scheming
As grader is classified, it is divided into reading area and non-reading area;
S3, extraction water meter image multi-channel feature, described multi-channel feature includes gradient orientation histogram, gradient magnitude, LUV face
Color characteristic, greyscale color feature, calculate the characteristic-integration figure of water meter image;
S4, travel through whole sliding window, utilize characteristic-integration figure to calculate each sliding window feature, utilize the Image Classifier pair that S2 trains
Sliding window is classified, and obtains target window;
S5, with extrapolation method estimate artwork characteristic-integration figure under multiple yardsticks, repeat S4, S5 step, obtain multiple dimensioned
Target window;
S6, rotation artwork, repeat S3, S4, S5, S6 step, obtain multidirectional target window.
Meter reading Region detection algorithms the most based on multi-feature fusion, it is characterised in that described step
Rapid S1 particularly as follows:
S1.1, by the water meter image pattern in RGB camera collection actual scene;
S1.2, the meter reading region in water meter image pattern acquired in S1.1 is carried out artificial mark, including water meter
The center of reading area (x, y), length h, width w and angle a.
Meter reading Region detection algorithms the most based on multi-feature fusion, it is characterised in that described water
Table image pattern includes following different parameter: illumination, visual angle, water meter type, the water meter extent of damage.
Meter reading Region detection algorithms the most based on multi-feature fusion, it is characterised in that described step
Rapid S2 particularly as follows:
S2.1, mark according to step S1 gained water meter image and reading area, cut out reading area and non-reading, for target
Classification;
S2.2, the multi-channel feature of extraction S2.1 institute cutting image, with multi-channel feature for input, train integrated decision tree classification
Whether device is to classify in reading area to this region.
Meter reading Region detection algorithms the most based on multi-feature fusion, it is characterised in that described step
Rapid S3 particularly as follows:
S3.1, extraction water meter image multi-channel feature, including gradient orientation histogram, gradient magnitude, LUV color characteristic, gray scale
Color characteristic, and calculate characteristic-integration figure;
The computational methods of described gradient direction:
Wherein (i j) is pixel (i, j) angle of place's gradient direction to O;I is image, and x represents horizontal direction, and y represents Vertical Square
To, i represents pixel coordinate in the horizontal direction;J represents pixel coordinate in vertical direction;
The computational methods of described gradient magnitude:
Wherein (i j) is the gradient magnitude at pixel (i.j) place to M;I is image, and x represents horizontal direction, and y represents vertical direction, i
Represent pixel coordinate in the horizontal direction;J represents pixel coordinate in vertical direction.
Meter reading Region detection algorithms the most based on multi-feature fusion, it is characterised in that described step
Rapid S4 particularly as follows:
S4.1, travel through each sliding window, utilize step S3 gained characteristic-integration figure, calculate the multi-channel feature of each sliding window, go forward side by side
Row Feature Fusion;
S4.2, with S4.1 merge characteristic vector for input, utilize S2 training gained Image Classifier sliding window is classified,
Obtain sliding window significance;
S4.3, the target detected is carried out maximization suppression, obtain detect target.
Meter reading Region detection algorithms the most based on multi-feature fusion, it is characterised in that described step
Rapid S5 particularly as follows:
S5.1, utilize the statistical property of adjacent scalogram picture, by the method for extrapolation estimate multiple dimensioned under artwork feature, extrapolation
Algorithm is as follows:
The wherein yardstick of S representative image, CsRepresenting the characteristics of image under yardstick S, (C, s) characteristic pattern C carries out yardstick is S to R
Resampling;λ is the channel parameters that the statistical analysis by image obtains;Ω represents different image channels;Extrapolation algorithm passes through
The characteristics of image calculated under a certain fixed size quickly calculates the characteristics of image of multiple adjacent yardstick;
S5.2, repeat S4, S5, obtain multiple dimensioned under reading area target window.
Meter reading Region detection algorithms the most based on multi-feature fusion, it is characterised in that described step
Rapid S6 particularly as follows:
With 5 degree for interval, multiple rotary original water gauge image, repeats S3, S4, S5, S6 step, is calculated multidirectional water meter reading
Number region, i.e. obtains multidirectional target window.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610513983.8A CN106156771B (en) | 2016-06-30 | 2016-06-30 | water meter reading area detection algorithm based on multi-feature fusion |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610513983.8A CN106156771B (en) | 2016-06-30 | 2016-06-30 | water meter reading area detection algorithm based on multi-feature fusion |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106156771A true CN106156771A (en) | 2016-11-23 |
CN106156771B CN106156771B (en) | 2020-01-31 |
Family
ID=57350987
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610513983.8A Active CN106156771B (en) | 2016-06-30 | 2016-06-30 | water meter reading area detection algorithm based on multi-feature fusion |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106156771B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111832328A (en) * | 2019-04-15 | 2020-10-27 | 北京京东尚科信息技术有限公司 | Bar code detection method, bar code detection device, electronic equipment and medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102842045A (en) * | 2012-08-03 | 2012-12-26 | 华侨大学 | Pedestrian detection method based on combined features |
US20130070099A1 (en) * | 2011-09-20 | 2013-03-21 | Honeywell International Inc. | Image based dial gauge reading |
CN103209211A (en) * | 2013-03-06 | 2013-07-17 | 江苏运赢物联网产业发展有限公司 | Meter reading recognizing method, meter self-service reading system and reading method thereof |
CN103218627A (en) * | 2013-01-31 | 2013-07-24 | 沈阳航空航天大学 | Image detection method and device |
-
2016
- 2016-06-30 CN CN201610513983.8A patent/CN106156771B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20130070099A1 (en) * | 2011-09-20 | 2013-03-21 | Honeywell International Inc. | Image based dial gauge reading |
CN102842045A (en) * | 2012-08-03 | 2012-12-26 | 华侨大学 | Pedestrian detection method based on combined features |
CN103218627A (en) * | 2013-01-31 | 2013-07-24 | 沈阳航空航天大学 | Image detection method and device |
CN103209211A (en) * | 2013-03-06 | 2013-07-17 | 江苏运赢物联网产业发展有限公司 | Meter reading recognizing method, meter self-service reading system and reading method thereof |
Non-Patent Citations (5)
Title |
---|
DONGDONG WANG 等: "A Saliency-Based Cascade Method for Fast Traffic Sign Detection", 《2015 IEEE INTELLIGENT VEHICLES SYMPOSIUM》 * |
吴涛: "《图像分割的认知物理学方法》", 30 April 2015, 北京:中国水利水电出版社 * |
朱峰: "基于数据流和精确定位的多线程行人探测系统", 《视频应用与工程》 * |
肖金秀 等: "《多媒体技术及应用》", 31 December 2004, 北京:冶金工业出版社 * |
赵池航 等: "《交通信息感知理论与方法》", 30 September 2014, 南京:东南大学出版社 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111832328A (en) * | 2019-04-15 | 2020-10-27 | 北京京东尚科信息技术有限公司 | Bar code detection method, bar code detection device, electronic equipment and medium |
Also Published As
Publication number | Publication date |
---|---|
CN106156771B (en) | 2020-01-31 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Qin et al. | Object-based 3-D building change detection on multitemporal stereo images | |
US9846946B2 (en) | Objection recognition in a 3D scene | |
Huang et al. | A new building extraction postprocessing framework for high-spatial-resolution remote-sensing imagery | |
Wen et al. | A novel automatic change detection method for urban high-resolution remotely sensed imagery based on multiindex scene representation | |
CN109255317B (en) | Aerial image difference detection method based on double networks | |
Guan et al. | Iterative tensor voting for pavement crack extraction using mobile laser scanning data | |
Marin et al. | Learning appearance in virtual scenarios for pedestrian detection | |
Ke et al. | A review of methods for automatic individual tree-crown detection and delineation from passive remote sensing | |
CN102043945B (en) | License plate character recognition method based on real-time vehicle tracking and binary index classification | |
CN109598794B (en) | Construction method of three-dimensional GIS dynamic model | |
Awrangjeb et al. | Building detection in complex scenes thorough effective separation of buildings from trees | |
US9639748B2 (en) | Method for detecting persons using 1D depths and 2D texture | |
CN104200521B (en) | High Resolution SAR Images building target three-dimensional rebuilding method based on model priori | |
CN103617426B (en) | Pedestrian target detection method under interference by natural environment and shelter | |
CN104077577A (en) | Trademark detection method based on convolutional neural network | |
CN106934386B (en) | A kind of natural scene character detecting method and system based on from heuristic strategies | |
CN102214309B (en) | Special human body recognition method based on head and shoulder model | |
CN104978567B (en) | Vehicle checking method based on scene classification | |
CN106610969A (en) | Multimodal information-based video content auditing system and method | |
CN101981582A (en) | Method, apparatus, and program for detecting object | |
CN103390164A (en) | Object detection method based on depth image and implementing device thereof | |
Shahab et al. | How salient is scene text? | |
CN105279772A (en) | Trackability distinguishing method of infrared sequence image | |
CN113469097B (en) | Multi-camera real-time detection method for water surface floaters based on SSD network | |
Lambers et al. | Towards detection of archaeological objects in high-resolution remotely sensed images: the Silvretta case study |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |