CN109285151A - A kind of AI intelligent dimming method - Google Patents
A kind of AI intelligent dimming method Download PDFInfo
- Publication number
- CN109285151A CN109285151A CN201811109672.0A CN201811109672A CN109285151A CN 109285151 A CN109285151 A CN 109285151A CN 201811109672 A CN201811109672 A CN 201811109672A CN 109285151 A CN109285151 A CN 109285151A
- Authority
- CN
- China
- Prior art keywords
- value
- image
- roi
- region
- intelligent dimming
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- H—ELECTRICITY
- H05—ELECTRIC TECHNIQUES NOT OTHERWISE PROVIDED FOR
- H05B—ELECTRIC HEATING; ELECTRIC LIGHT SOURCES NOT OTHERWISE PROVIDED FOR; CIRCUIT ARRANGEMENTS FOR ELECTRIC LIGHT SOURCES, IN GENERAL
- H05B45/00—Circuit arrangements for operating light-emitting diodes [LED]
- H05B45/10—Controlling the intensity of the light
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30168—Image quality inspection
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02B—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO BUILDINGS, e.g. HOUSING, HOUSE APPLIANCES OR RELATED END-USER APPLICATIONS
- Y02B20/00—Energy efficient lighting technologies, e.g. halogen lamps or gas discharge lamps
- Y02B20/40—Control techniques providing energy savings, e.g. smart controller or presence detection
Abstract
The present invention relates to intelligent dimming technical fields, refer in particular to AI intelligent dimming method, comprising the following steps: step 1: pre-processing to image, are effectively filtered for since actual environment influences generation picture noise;Step 2: treating polishing object by using deep learning and machine vision mode and carry out automatic identification, obtain polishing region of interest ROI;Step 3: to carrying out feature extraction to polishing object edge and texture in ROI and statistically analyze, in conjunction with human eye characteristic, the mathematical model for establishing a non-reference picture quality evaluation, by carrying out objective evaluation to picture quality, whether checking image quality is clear, uniform;Step 4: by local dimming, obtaining best GiLight value.The present invention automatically analyzes region of interest ROI using deep learning and machine vision method, realizes that the analysis of ROI region image intelligent is AI image quality measure, obtains the best G in each regioniLight value, to control each channel lamplight brightness.
Description
Technical field
The present invention relates to intelligent dimming technical fields, refer in particular to a kind of AI intelligent dimming method.
Background technique
Industrial products are many kinds of, in the production process of industrial products, need to carry out observation detection to product, that is, need
Obtain the image information of product.And the principal element for influencing image definition, details performance and gray-level performance is polishing.And
Currently used polishing mode is manually to get to light source by polishing object, but such mode is by artificial master
Viewing is rung greatly and otherness is larger, and the quality of polishing cannot be guaranteed unification, inefficiency, therefore, using a kind of automatical and efficient
Polishing mode be people's urgent problem to be solved.
Summary of the invention
The present invention provides a kind of AI intelligent dimming method for the above technical issues.
In order to solve the above-mentioned technical problem, the present invention adopts the following technical scheme:
A kind of AI intelligent dimming method, comprising the following steps:
Step 1: image being pre-processed, carries out effective mistake for since actual environment influences the picture noise generated
Filter;
Step 2: treating polishing object by using the mode of deep learning and machine vision and carry out automatic identification, obtain and beat
The region of interest ROI of light;
Step 3: feature extraction being carried out to the edge and texture to polishing object in ROI and is statisticallyd analyze, in conjunction with human eye
Distinctive feature establishes the mathematical model of a non-reference picture quality evaluation, by carrying out objective evaluation to picture quality, examines
Whether picture quality is clear, uniform;
Step 4: by local dimming, obtaining best GiLight value.
Further, step 3 specifically includes: objective evaluation is carried out to picture quality by following formula:
Wherein roiMSE indicates the mean square error of image gradient, and W, H are the width and height to polishing object in ROI region,
GmeanFor the average gray value on present image gradient locations (i, j), WijFor the gray value of (i, j) in sliding window;roiPSNR
For Y-PSNR, the i.e. ratio of peak signal amount L and noise intensity;RoiDIS indicates the line in ROI region to polishing object
Discrete feature is managed, wherein Bi、Bmean、BmaxFor current value, mean value, the maximum for carrying out gray-scale watermark statistics in unit window
Value.
Then following image quality measure function is proposed:
Score=W1×roiPSNR+W2×mean(Sobel(img))+W3×roiDIS;
Wherein W1、W2And W3It is particular value, passes through certain amount sample analysis and the warp obtained in conjunction with human eye distinctive feature
Value is tested, G is found outmax。
Further, step 4 specifically includes:
Step 4.1: by global light, calculating max (Score) and obtain the best gradient of image, and obtain corresponding GmaxAnd
roiPSNRg,roiDISg;
Step 4.2: area maps being carried out according to LED light source characteristic, roiPSNRi, the roiDISi for calculating each region are obtained
G outi;
Step 4.3: controlling each channel G respectivelyiValue, is intended to Gmax, step 4.2 and step 4.3 are iterated to calculate, is calculated
Loss1=fabs (roiPSNRi-roiPSNRg), Loss2=fabs (roiDISi-roiDISg) out;
Step 4.4: solving min (Loss1+Loss2), acquire the best G in each channeliLight value.
Further, step 1 specifically includes: RGB channel separates gaussian filtering and connected domain filtering.
Further, horizontally and vertically for the region of interest ROI of polishing, it is calculated by gradient analysis
Gradient both horizontally and vertically, to detect gaussian filtering and whether the filtered image of connected domain is clear.
Beneficial effects of the present invention: the present invention is without setting reference picture, using the method for deep learning and machine vision
Region of interest ROI is automatically analyzed, realizes that intelligence is AI image quality measure, obtains the best G in each regioniLight value, from
And control the luminosity that each channel adjusts corresponding light.
Detailed description of the invention
Fig. 1 is flow chart of the method for the present invention.
Specific embodiment
For the ease of the understanding of those skilled in the art, the present invention is made further below with reference to embodiment and attached drawing
Bright, the content that embodiment refers to not is limitation of the invention.The present invention is described in detail below in conjunction with attached drawing.
Refering to Figure 1, a kind of AI intelligent dimming method provided by the invention, comprising the following steps:
Step 1: image being pre-processed, carries out effective mistake for since actual environment influences the picture noise generated
Filter;
Step 2: treating polishing object by using the mode of deep learning and machine vision and carry out automatic identification, obtain and beat
The region of interest ROI of light;
Step 3: feature extraction being carried out to the edge and texture to polishing object in ROI and is statisticallyd analyze, in conjunction with human eye
Distinctive feature establishes the mathematical model of a non-reference picture quality evaluation, by carrying out objective evaluation to picture quality, examines
Whether picture quality is clear, uniform;
Step 4: by local dimming, obtaining best GiValue.
The present invention automatically analyzes area-of-interest using the method for deep learning and machine vision without setting reference picture
ROI realizes that intelligence is AI image quality measure, obtains the best G in each regioniLight value, to control each channel adjustment
The luminosity of corresponding light.
Specific automatic identification mode includes carrying out image object point using the method for machine vision for common object
It cuts, object detection is carried out using deep learning method for some specific objects, specific deep learning method includes but not
It is defined in the neural networks such as VGG, resNet.
A kind of AI intelligent dimming method, step 3 described in the present embodiment specifically include:
Objective evaluation is carried out to picture quality by following formula:
Wherein roiMSE indicates the mean square error of image gradient, and W, H are the width and height to polishing object in ROI region,
GmeanFor the average gray value on present image gradient locations (i, j), WijFor the gray value of (i, j) in sliding window;roiPSNR
For Y-PSNR, the i.e. ratio of peak signal amount L and noise intensity;RoiDIS indicates the line in ROI region to polishing object
Discrete feature is managed, wherein Bi、Bmean、BmaxFor current value, mean value, the maximum for carrying out gray-scale watermark statistics in unit window
Value.
Then following image quality measure function is proposed:
Score=W1×roiPSNR+W2×mean(Sobel(img))+W3×roiDIS;
Wherein W1、W2And W3It is particular value, passes through certain amount sample analysis and the warp obtained in conjunction with human eye distinctive feature
Value is tested, G is found outmax.Specifically, wherein W1It is 0.62, W2It is 0.24, W3It is 0.14.
Wherein, the MSE is mean squared error, and the PSNR is that peak signal to noise, AI are
Artificial intelligence Artificial Intelligence, GmaxFor the maximum value in calculated G value, G value is light value.
The sliding window is 3 × 3 neighborhoods, and peak signal amount is single channel image maximum gradation value L=255.
A kind of AI intelligent dimming method, step 4 described in the present embodiment specifically include:
Step 4.1: by global light, calculating max (Score) and obtain the best gradient of image, and obtain corresponding GmaxAnd
roiPSNRg,roiDISg;
Step 4.2: area maps being carried out according to LED light source characteristic, roiPSNRi, the roiDISi for calculating each region are obtained
G outi;
Step 4.3: controlling each channel G respectivelyiValue, is intended to Gmax, step 4.2 and step 4.3 are iterated to calculate, is calculated
Loss1=fabs (roiPSNRi-roiPSNRg), Loss2=fabs (roiDISi-roiDISg) out;
Step 4.4: solving min (Loss1+Loss2), acquire the best G in each channeliLight value.
Wherein roiPSNRi and GiIn i indicate ith zone, GiValue refers to the G value in ith zone, roiPSNRi
For the Y-PSNR of ith zone.
A kind of AI intelligent dimming method, step 1 described in the present embodiment specifically include: RGB channel separate gaussian filtering and
Connected domain filtering.
Specifically, the gaussian filtering is referred to an each of template (or convolution, mask) scan image
Pixel goes the value of alternate template central pixel point with the weighted average gray value of pixel in the determining neighborhood of template, for effective
The picture noise that ground filtering is introduced due to illumination, light modulator, camera.
Specifically, the picture noise that connected domain filtering effectively filtering is introduced due to illumination, light modulator, camera.
A kind of AI intelligent dimming method described in the present embodiment, for polishing region of interest ROI horizontal direction and
Vertical direction calculates gradient both horizontally and vertically by gradient analysis, that is, Sobel method, to detect gaussian filtering and company
Whether the logical filtered image in domain is clear.
The above is only present pre-ferred embodiments, is not intended to limit the present invention in any form, although
The present invention is disclosed as above with preferred embodiment, and however, it is not intended to limit the invention, any person skilled in the art,
It does not depart within the scope of technical solution of the present invention, when the technology contents using the disclosure above make a little change or are modified to equivalent change
The equivalent embodiment of change, but without departing from the technical solutions of the present invention, technology refers to above embodiments according to the present invention
Made any simple modification, equivalent change and modification, belong in the range of technical solution of the present invention.
Claims (5)
1. a kind of AI intelligent dimming method, it is characterised in that: the following steps are included:
Step 1: image being pre-processed, is effectively filtered for the picture noise for influencing to generate due to actual environment;
Step 2: treating polishing object by using the mode of deep learning and machine vision and carry out automatic identification, obtain polishing
Region of interest ROI;
Step 3: feature extraction being carried out to the edge and texture to polishing object in ROI and is statisticallyd analyze, is distinguished in conjunction with human eye
Characteristic establishes the mathematical model of a non-reference picture quality evaluation, by carrying out objective evaluation, checking image to picture quality
Whether quality is clear, uniform;
Step 4: by local dimming, obtaining best GiLight value.
2. a kind of AI intelligent dimming method according to claim 1, it is characterised in that:
Step 3 specifically includes: objective evaluation is carried out to picture quality by following formula:
Wherein roiMSE indicates the mean square error of image gradient, and W, H are the width and height to polishing object in ROI region, GmeanFor
Average gray value on present image gradient locations (i, j), WijFor the gray value of (i, j) in sliding window;RoiPSNR is peak value
Signal-to-noise ratio, the i.e. ratio of peak signal amount L and noise intensity;RoiDIS indicates that the texture in ROI region to polishing object is discrete
Characteristic, wherein Bi、Bmean、BmaxFor current value, mean value, the maximum value for carrying out gray-scale watermark statistics in unit window.
Then following image quality measure function is proposed:
Score=W1×roiPSNR+W2×mean(Sobel(img))+W3×roiDIS;
Wherein W1、W2And W3It is particular value, passes through certain amount sample analysis and the experience obtained in conjunction with human eye distinctive feature
Value, finds out Gmax。
3. a kind of AI intelligent dimming method according to claim 2, it is characterised in that:
Step 4 specifically includes:
Step 4.1: by global light, calculating max (Score) and obtain the best gradient of image, and obtain corresponding GmaxAnd
roiPSNRg,roiDISg;
Step 4.2: area maps being carried out according to LED light source characteristic, roiPSNRi, the roiDISi for calculating each region obtain Gi;
Step 4.3: controlling each channel G respectivelyiValue, is intended to Gmax, step 4.2 and step 4.3 are iterated to calculate, is calculated
Loss1=fabs (roiPSNRi-roiPSNRg), Loss2=fabs (roiDISi-roiDISg);
Step 4.4: solving min (Loss1+Loss2), acquire the best G in each channeliLight value.
4. a kind of AI intelligent dimming method according to claim 1, it is characterised in that: step 1 specifically includes: RGB channel
Separate gaussian filtering and connected domain filtering.
5. a kind of AI intelligent dimming method according to claim 4, it is characterised in that: for the area-of-interest of polishing
ROI horizontally and vertically, by gradient analysis calculates gradient both horizontally and vertically, to detect gaussian filtering
It is whether clear with the filtered image of connected domain.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811109672.0A CN109285151B (en) | 2018-09-21 | 2018-09-21 | AI intelligent dimming method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811109672.0A CN109285151B (en) | 2018-09-21 | 2018-09-21 | AI intelligent dimming method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109285151A true CN109285151A (en) | 2019-01-29 |
CN109285151B CN109285151B (en) | 2022-04-12 |
Family
ID=65182029
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811109672.0A Active CN109285151B (en) | 2018-09-21 | 2018-09-21 | AI intelligent dimming method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109285151B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111707455A (en) * | 2020-07-03 | 2020-09-25 | 深圳爱克莱特科技股份有限公司 | Smooth dimming method and system for lamp |
CN113189113A (en) * | 2021-04-30 | 2021-07-30 | 聚时科技(上海)有限公司 | Intelligent digital light source and method based on visual detection |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102466945A (en) * | 2010-11-19 | 2012-05-23 | 北京海鑫智圣技术有限公司 | LED supplementary lighting and image clipping evaluation system in standard image acquisition device |
US20180068195A1 (en) * | 2016-09-07 | 2018-03-08 | Apple, Inc. | Multi-Dimensional Objective Metric Concentering |
CN107959767A (en) * | 2017-12-14 | 2018-04-24 | 中国科学院长春光学精密机械与物理研究所 | A kind of focusing light-dimming method using TV track result as guiding |
-
2018
- 2018-09-21 CN CN201811109672.0A patent/CN109285151B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102466945A (en) * | 2010-11-19 | 2012-05-23 | 北京海鑫智圣技术有限公司 | LED supplementary lighting and image clipping evaluation system in standard image acquisition device |
US20180068195A1 (en) * | 2016-09-07 | 2018-03-08 | Apple, Inc. | Multi-Dimensional Objective Metric Concentering |
CN107959767A (en) * | 2017-12-14 | 2018-04-24 | 中国科学院长春光学精密机械与物理研究所 | A kind of focusing light-dimming method using TV track result as guiding |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111707455A (en) * | 2020-07-03 | 2020-09-25 | 深圳爱克莱特科技股份有限公司 | Smooth dimming method and system for lamp |
CN111707455B (en) * | 2020-07-03 | 2021-08-13 | 深圳爱克莱特科技股份有限公司 | Smooth dimming method and system for lamp |
CN113189113A (en) * | 2021-04-30 | 2021-07-30 | 聚时科技(上海)有限公司 | Intelligent digital light source and method based on visual detection |
Also Published As
Publication number | Publication date |
---|---|
CN109285151B (en) | 2022-04-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Dhara et al. | Color cast dependent image dehazing via adaptive airlight refinement and non-linear color balancing | |
CN108596849B (en) | Single image defogging method based on sky region segmentation | |
CN109635758B (en) | Intelligent building site video-based safety belt wearing detection method for aerial work personnel | |
CN101304489B (en) | Automatic exposure method and apparatus | |
CN110570360B (en) | Retinex-based robust and comprehensive low-quality illumination image enhancement method | |
US20080232650A1 (en) | Face region detecting device, method, and computer readable recording medium | |
US11700457B2 (en) | Flicker mitigation via image signal processing | |
CN105654437A (en) | Enhancement method for low-illumination image | |
CN109269951A (en) | Floating tail-coal ash content, concentration, coarse granule detection method of content based on image | |
CN113109348B (en) | Paddle image transfer printing defect identification method based on machine vision | |
CN106778534B (en) | Method for identifying ambient light during vehicle running | |
CN103914708A (en) | Food variety detection method and system based on machine vision | |
CN106339994B (en) | A kind of image enchancing method | |
CN110288618B (en) | Multi-target segmentation method for uneven-illumination image | |
CN108267455A (en) | plastic film printing character defect detecting device and method | |
CN108921857A (en) | A kind of video image focus area dividing method towards monitoring scene | |
CN109285151A (en) | A kind of AI intelligent dimming method | |
CN107085707A (en) | A kind of license plate locating method based on Traffic Surveillance Video | |
CN111476732B (en) | Image fusion and denoising method and system | |
Liu et al. | Single image haze removal via depth-based contrast stretching transform | |
CN110022426A (en) | A kind of camera Intelligent supplemental lighting method based on image recognition | |
CN108765316B (en) | Mist concentration self-adaptive judgment method | |
CN104573743B (en) | A kind of facial image detection filter method | |
CN116843681B (en) | Intelligent detection method for coloring quality of carton surface | |
CN108711139B (en) | One kind being based on defogging AI image analysis system and quick response access control method |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |