CN106156771B - water meter reading area detection algorithm based on multi-feature fusion - Google Patents
water meter reading area detection algorithm based on multi-feature fusion Download PDFInfo
- Publication number
- CN106156771B CN106156771B CN201610513983.8A CN201610513983A CN106156771B CN 106156771 B CN106156771 B CN 106156771B CN 201610513983 A CN201610513983 A CN 201610513983A CN 106156771 B CN106156771 B CN 106156771B
- Authority
- CN
- China
- Prior art keywords
- water meter
- image
- reading area
- feature
- scale
- 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
Images
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
The invention discloses water meter reading area detection algorithms based on multi-feature fusion, which comprises the following steps of S1 obtaining training data, S2 cutting a reading area and a non-reading area in a water meter image, extracting multi-channel features of the cutting area, performing feature fusion, and inputting the features into a training target classifier, S3 extracting the multi-channel features of the water meter image, calculating a feature integral diagram, S4 calculating the features of each sliding window by using the feature integral diagram, inputting the fusion features, classifying the sliding windows by using the classifier obtained by S2 training, obtaining target windows, S5 estimating the feature diagrams of the original image under multiple scales by using an extrapolation method, repeating S4 and S5, obtaining multi-scale target windows, S6 rotating the original image, repeating S3, S4, S5 and S6, obtaining the target windows, and providing multi-direction water meter reading area detection algorithms which are accurate, robust and practical.
Description
Technical Field
The invention relates to the field of artificial intelligence, in particular to water meter reading area detection algorithms based on multi-feature fusion.
Background
In recent years, with the development of mobile internet and the popularization of digital products, image data from different devices (smart phones, digital cameras, even cameras such as automatic street view cars and unmanned aerial vehicles) are continuously and explosively increased, parts of image data in the massive images carry text information, and the text information usually contains very beneficial semantic information, for example, the text information can be descriptions of buildings, shops, traffic signs, road signs, commodity names and the like, therefore, the high-level semantic information can be widely applied to occasions such as machine reading, automatic interpretation, image retrieval, video retrieval, language translation, automatic driving, robot navigation and the like by , intelligent visual text analysis technologies are more urgently needed by human beings, technologies for extracting and understanding the text information from the perspective of machine vision, the visual text analysis relates to series of science and science, such as image processing, mode recognition, computer vision, machine learning and psychology, and the like, and is an important research direction of relevant fields in .
The automatic reading of the water meter based on computer vision is important applications in visual character analysis, and can replace the existing manual water meter reading mode, so that the water meter reading becomes an automatic process.
The problem to be solved primarily for the vision-based automatic reading of the water meter is the detection of a reading area, and the current mainstream method is an image processing-based method, wherein the reading area is determined through the steps of image denoising, image binarization based on color characteristics, direction detection based on line detection, area segmentation and the like. However, the method has poor adaptability to conditions such as illumination, deformation and masking in various complex scenes, is easy to interfere and has poor robustness.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides water meter reading area detection algorithms based on multi-feature fusion.
The purpose of the invention is realized by the following technical scheme:
A water meter reading area detection algorithm based on multi-feature fusion, comprising the following steps:
s1, acquiring training data, shooting a water meter image sample through a camera, manually marking a water meter reading area in the image, and acquiring the central position, length and width information of the water meter reading;
s2, cutting a reading area and a non-reading area in the water meter image, extracting multi-channel characteristics of the cutting area, performing characteristic fusion, and inputting the extracted multi-channel characteristics into a training image classifier; the cutting area is an uncertain area, and is classified by an image classifier into a reading area and a non-reading area;
s3, extracting multi-channel characteristics of the water meter image, wherein the multi-channel characteristics comprise a gradient direction histogram, a gradient amplitude, LUV color characteristics and gray color characteristics, and calculating a characteristic integral chart of the water meter image;
s4, traversing all sliding windows, calculating the characteristics of each sliding window by using the characteristic integral graph, and classifying the sliding windows by using the image classifier trained in S2 to obtain a target window;
s5, estimating the feature integral graph of the original image under multiple scales by an extrapolation method, repeating the steps S4 and S5, and acquiring a multi-scale target window;
and S6, rotating the original graph, repeating the steps S3, S4, S5 and S6, and acquiring a multi-directional target window.
The step S1 specifically includes:
s1.1, acquiring a water meter image sample in an actual scene through an RGB camera;
s1.2, artificially marking a water meter reading area in the water meter image sample acquired in the S1.1, wherein the water meter reading area comprises the center position (x, y), the length h, the width w and the angle a of the water meter reading area.
The water meter image samples include the following different parameters: illumination, viewing angle, water meter type, degree of water meter damage. This is done to ensure sample diversity.
The step S2 specifically includes:
s2.1, according to the water meter image and the reading area label obtained in the step S1, cutting a reading area and a non-reading area for target classification;
and S2.2, extracting multi-channel characteristics of the image cut in the S2.1, and training an integrated decision tree classifier to classify whether the region is a reading region or not by taking the multi-channel characteristics as input.
The step S3 specifically includes:
s3.1, extracting multi-channel characteristics of the water meter image, including a gradient direction histogram, a gradient amplitude, LUV color characteristics and gray color characteristics, and calculating a characteristic integral chart;
wherein O (i, j) is the angle of the gradient direction at the pixel point (i, j); i is an image, x represents the horizontal direction, y represents the vertical direction, and I represents the coordinate of a pixel point in the horizontal direction; j represents the coordinate of the pixel point in the vertical direction;
wherein M (i, j) is the gradient magnitude at the pixel point (i.j); i is an image, x represents the horizontal direction, y represents the vertical direction, and I represents the coordinate of a pixel point in the horizontal direction; j represents the coordinate of the pixel point in the vertical direction.
The step S4 specifically includes:
s4.1, traversing each sliding window, calculating the multichannel characteristics of each sliding window by using the characteristic integral graph obtained in the step S3, and performing characteristic fusion;
s4.2, classifying the sliding window by using the image classifier obtained by training S2 by taking the feature vector fused in S4.1 as input to obtain the significance of the sliding window;
and S4.3, carrying out maximization inhibition on the detected target to obtain the detected target.
The step S5 specifically includes:
s5.1, estimating original image characteristics under multiple scales by using statistical characteristics of adjacent scale images through an extrapolation method, wherein an extrapolation algorithm is as follows:
wherein S represents the scale of the image, CsRepresenting the image characteristics under the scale S, wherein R (C, S) is the resampling with the scale S for the characteristic image C, lambda is a channel parameter obtained by the statistical analysis of a large number of images, omega represents different image channels, and an extrapolation algorithm rapidly calculates the image characteristics of a plurality of adjacent scales through the image characteristics calculated under a certain fixed scale;
s5.2, repeating the steps S4 and S5 to obtain a reading area target window under multiple scales.
The step S6 specifically includes:
and (4) rotating the original water meter image for multiple times at intervals of 5 degrees, repeating the steps S3, S4, S5 and S6, and calculating to obtain a multi-directional water meter reading area, namely acquiring a multi-directional target window.
Compared with the prior art, the invention has the following advantages and beneficial effects:
1. the invention utilizes the multi-scale feature extraction method of multi-feature fusion to efficiently extract rich information under multiple scales, solves the first difficult problem in automatic reading of the water meter, namely multi-scale and multidirectional water meter reading area detection, and has great application prospect.
2. In addition, the invention adopts a multi-feature fusion method, so the algorithm has strong robustness and good adaptability to various complex scenes.
Drawings
Fig. 1 is a flow chart of water meter reading area detection algorithms based on multi-feature fusion according to the present invention.
Fig. 2 is a flow chart of the training of the algorithm described in fig. 1.
Detailed Description
The present invention will be described in further detail with reference to examples and drawings, but the embodiments of the present invention are not limited thereto.
As shown in fig. 1, the multi-feature fusion based water meter reading area detection algorithm mainly includes the following steps:
s1, acquiring training data; the training process is shown in fig. 2;
s1.1, collecting a large number of water meter image samples in an actual scene through an RGB camera, wherein the water meter image samples comprise various illumination, visual angles, water meter types, water meter damage degrees and the like, so that the diversity of the samples is ensured;
s1.2, artificially marking a water meter reading area in the water meter image sample acquired in the S1.1, wherein the water meter reading area comprises a central position (x, y), a length (h), a width (w) and an angle (a);
s2 train the target classifier:
s2.1, marking the water meter image and the reading area obtained in the step S1, and cutting out a reading area and a non-reading area for classifying targets;
s2.2, extracting multi-channel characteristics of the image cut in the S2.1, and training an integrated decision tree classifier to classify whether the region is a reading region or not by taking the characteristics as input;
s3 calculates image multi-channel features:
s3.1, extracting multi-channel characteristics of the water meter image, including a gradient direction histogram, a gradient amplitude, LUV color characteristics and gray color characteristics, and calculating a characteristic integral chart:
gradient amplitude calculation method:
s4 sliding window scan:
s4.1, traversing each sliding window, calculating the multichannel characteristics of each sliding window by using the characteristic integral graph obtained in the S3.1, and performing characteristic fusion;
s4.2, classifying the sliding window by using the classifier obtained by training S2 by taking the feature vector fused in S4.1 as input to obtain the significance of the sliding window;
and S4.3, carrying out maximization inhibition on the detected target to obtain the detected target.
S5 extrapolating the multi-scale features:
s5.1, estimating original image characteristics under multiple scales by using statistical characteristics of adjacent scale images through an extrapolation method, wherein an extrapolation algorithm is as follows:
wherein S represents the scale of the image, CsRepresenting image features at scale S, R (C, S) being features of opposite orderThe characteristic graph C is resampled with the scale S, lambda is a channel parameter obtained through the statistical analysis of the image, omega represents different image channels, and an extrapolation algorithm quickly calculates the image characteristics of a plurality of adjacent scales through the image characteristics calculated under a certain fixed scale;
s5.2, repeating the steps S4 and S5 to obtain a reading area target window under multiple scales.
S6 rotation detection:
and S6.1, rotating the original water meter image for multiple times at intervals of 5 degrees, and repeating the steps of S3, S4, S5 and S6 to obtain a multi-directional water meter reading area through calculation.
The above embodiments are preferred embodiments of the present invention, but the present invention is not limited to the above embodiments, and any other changes, modifications, substitutions, combinations, and simplifications which do not depart from the spirit and principle of the present invention should be construed as equivalents thereof, and all such changes, modifications, substitutions, combinations, and simplifications are intended to be included in the scope of the present invention.
Claims (7)
- S1, acquiring training data, shooting a water meter image sample through a camera, manually marking a water meter reading area in the image, and acquiring the information of the center position, the length and the width of the water meter reading;s2, cutting a reading area and a non-reading area in the water meter image, extracting multi-channel characteristics of the cutting area, performing characteristic fusion, and inputting the extracted multi-channel characteristics into a training image classifier; the cutting area is an uncertain area, and is classified by an image classifier into a reading area and a non-reading area;s3, extracting multi-channel characteristics of the water meter image, wherein the multi-channel characteristics comprise a gradient direction histogram, a gradient amplitude, LUV color characteristics and gray color characteristics, and calculating a characteristic integral chart of the water meter image;s4, traversing all sliding windows, calculating the characteristics of each sliding window by using the characteristic integral graph, and classifying the sliding windows by using the image classifier trained in S2 to obtain a target window;s5, estimating the feature integral graph of the original image under multiple scales by an extrapolation method, repeating the steps S4 and S5, and acquiring a multi-scale target window;s6, rotating the original drawing, repeating the steps S3, S4, S5 and S6, and acquiring a multi-directional target window;s4.1, traversing each sliding window, calculating the multichannel characteristics of each sliding window by using the characteristic integral graph obtained in the step S3, and performing characteristic fusion;s4.2, classifying the sliding window by using the image classifier obtained by training S2 by taking the feature vector fused in S4.1 as input to obtain the significance of the sliding window;and S4.3, carrying out maximization inhibition on the detected target to obtain the detected target.
- 2. The multi-feature fusion based water meter reading area detection algorithm of claim 1, wherein the step S1 specifically comprises: s1.1, acquiring a water meter image sample in an actual scene through an RGB camera;s1.2, artificially marking a water meter reading area in the water meter image sample acquired in the S1.1, wherein the water meter reading area comprises the center position (x, y), the length h, the width w and the angle a of the water meter reading area.
- 3. The multi-feature fusion based water meter reading area detection algorithm of claim 2, wherein the water meter image samples comprise the following different parameters: illumination, viewing angle, water meter type, degree of water meter damage.
- 4. The multi-feature fusion based water meter reading area detection algorithm of claim 1, wherein the step S2 specifically comprises: s2.1, according to the water meter image and the reading area label obtained in the step S1, cutting a reading area and a non-reading area for target classification;and S2.2, extracting multi-channel characteristics of the image cut in the S2.1, and training an integrated decision tree classifier to classify whether the region is a reading region or not by taking the multi-channel characteristics as input.
- 5. The multi-feature fusion-based water meter reading area detection algorithm as claimed in claim 1, wherein the step S3 is specifically that S3.1, extracting multi-channel features of the water meter image, including gradient direction histogram, gradient amplitude, LUV color feature, and gray color feature, and calculating a feature integral graph;the gradient direction calculation method comprises the following steps:wherein O (i, j) is the angle of the gradient direction at the pixel point (i, j); i is an image, x represents the horizontal direction, y represents the vertical direction, and I represents the coordinate of a pixel point in the horizontal direction; j represents the coordinate of the pixel point in the vertical direction;the gradient amplitude calculation method comprises the following steps:wherein M (i, j) is the gradient magnitude at the pixel point (i.j); i is an image, x represents the horizontal direction, y represents the vertical direction, and I represents the coordinate of a pixel point in the horizontal direction; j represents the coordinate of the pixel point in the vertical direction.
- 6. The algorithm for detecting the water meter reading area based on the multi-feature fusion as claimed in claim 1, wherein the step S5 is specifically that S5.1, the statistical characteristics of the adjacent scale images are used to estimate the original image features under the multi-scale by an extrapolation method, and the extrapolation algorithm is as follows:the method comprises the steps of obtaining a characteristic graph C, wherein S represents the scale of an image, Cs represents the image characteristic under the scale S, R (C, S) is resampling with the scale of S on the characteristic graph C, lambda is a channel parameter obtained through image statistical analysis, omega represents different image channels, and an extrapolation algorithm quickly calculates the image characteristics of a plurality of adjacent scales through the image characteristics calculated under a certain fixed scale;s5.2, repeating the steps S4 and S5 to obtain a reading area target window under multiple scales.
- 7. The algorithm for detecting the reading area of the water meter based on the multi-feature fusion as claimed in claim 1, wherein the step S6 is specifically that the original water meter image is rotated for a plurality of times at intervals of 5 degrees, and the steps S3, S4, S5 and S6 are repeated to calculate the multi-directional reading area of the water meter, that is, a multi-directional target window is obtained.
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 CN106156771A (en) | 2016-11-23 |
CN106156771B true 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) |
Families Citing this family (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 (3)
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 |
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 |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9135492B2 (en) * | 2011-09-20 | 2015-09-15 | Honeywell International Inc. | Image based dial gauge reading |
-
2016
- 2016-06-30 CN CN201610513983.8A patent/CN106156771B/en active Active
Patent Citations (3)
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 |
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 (2)
Title |
---|
Dongdong Wang 等.A Saliency-Based Cascade Method for Fast Traffic Sign Detection.《2015 IEEE Intelligent Vehicles Symposium》.2015,第180-185页. * |
基于数据流和精确定位的多线程行人探测系统;朱峰;《视频应用与工程》;20160531;第40卷(第5期);第121-128页 * |
Also Published As
Publication number | Publication date |
---|---|
CN106156771A (en) | 2016-11-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110175576B (en) | Driving vehicle visual detection method combining laser point cloud data | |
CN107424142B (en) | Weld joint identification method based on image significance detection | |
US9846946B2 (en) | Objection recognition in a 3D scene | |
CN109101924B (en) | Machine learning-based road traffic sign identification method | |
CN106651872B (en) | Pavement crack identification method and system based on Prewitt operator | |
CN107610114B (en) | optical satellite remote sensing image cloud and snow fog detection method based on support vector machine | |
CN109255317B (en) | Aerial image difference detection method based on double networks | |
CN108520226B (en) | Pedestrian re-identification method based on body decomposition and significance detection | |
Zhou et al. | Robust vehicle detection in aerial images using bag-of-words and orientation aware scanning | |
CN107392141B (en) | Airport extraction method based on significance detection and LSD (least squares distortion) line detection | |
CN110781836A (en) | Human body recognition method and device, computer equipment and storage medium | |
CN104077577A (en) | Trademark detection method based on convolutional neural network | |
CN111860439A (en) | Unmanned aerial vehicle inspection image defect detection method, system and equipment | |
CN108711172B (en) | Unmanned aerial vehicle identification and positioning method based on fine-grained classification | |
CN108509950B (en) | Railway contact net support number plate detection and identification method based on probability feature weighted fusion | |
CN110751619A (en) | Insulator defect detection method | |
CN114331986A (en) | Dam crack identification and measurement method based on unmanned aerial vehicle vision | |
KR20190059083A (en) | Apparatus and method for recognition marine situation based image division | |
CN112465854A (en) | Unmanned aerial vehicle tracking method based on anchor-free detection algorithm | |
CN108509826B (en) | Road identification method and system for remote sensing image | |
CN106156771B (en) | water meter reading area detection algorithm based on multi-feature fusion | |
CN116758421A (en) | Remote sensing image directed target detection method based on weak supervised learning | |
CN107704864A (en) | Well-marked target detection method based on image object Semantic detection | |
CN116188943A (en) | Solar radio spectrum burst information detection method and device | |
CN113689365B (en) | Target tracking and positioning method based on Azure Kinect |
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 |