Search

CN-122023544-A - LED color image analysis method and system based on multi-station acquisition

CN122023544ACN 122023544 ACN122023544 ACN 122023544ACN-122023544-A

Abstract

The invention discloses an LED color image analysis method and system based on multi-station acquisition, and relates to the technical field of LED image color analysis. The method comprises the following steps of S1, collecting multi-source optical data and standard LED color plate images of all stations, carrying out brightness normalization to form a corrected input image, S2, extracting color characteristics of a standard color area, carrying out color transformation on the corrected input image, S3, locating an LED luminous area, screening effective pixels to carry out stable aggregation to generate an LED area stable chromaticity vector, finishing indication labeling of target color types, S4, comprehensively evaluating long-term color drift indexes of the stations, determining compensation intensity and applying the compensation intensity pixel by pixel, and S5, carrying out structured feedback on monitoring data of all the stations and dynamic version management of compensation parameters. The problem of cross-station color drift inconsistency caused by the influence of optical deviation and environmental change on the color appearance of the LED under the conditions of multiple stations and multiple cameras is solved.

Inventors

  • YANG LIJIANG
  • YANG CHENGMENG
  • CHEN LIYUN

Assignees

  • 浙江海英俊电子科技有限公司

Dates

Publication Date
20260512
Application Date
20260120

Claims (10)

  1. 1. The LED color image analysis method based on multi-station acquisition is characterized by comprising the following steps of: S1, acquiring multi-source optical data and standard LED color board images of each station, weakening illumination differences of different stations through a brightness normalization method, and forming corrected input images with consistent illumination; S2, extracting color features of a standard color area from a standard LED color plate image, fitting to obtain a station color mapping matrix, and performing color transformation on a corrected input image pixel by pixel to obtain a standardized corrected image under a unified reference chromaticity space; S3, positioning an LED luminous area in the standardized correction image, screening and suppressing noise and local flicker through effective pixels, performing robust aggregation on chromaticity components of the effective pixels, generating an LED area robust chromaticity vector capable of resisting illumination and noise disturbance, and finishing indication labeling of target color types; S4, comprehensively analyzing the steady chromaticity vector of each station according to the color category, and evaluating the long-term color drift index of the station; s5, carrying out structured feedback and dynamic version management on the monitoring data of each station, and realizing standardized output of detection results and self-updating closed loop of parameters.
  2. 2. The multi-station acquisition-based LED color image analysis method according to claim 1, wherein the specific process of acquiring multi-source optical data and standard LED color plate images of each station and weakening illumination differences of different stations by a brightness normalization method is as follows: Acquiring multi-source optical data of each station in real time, wherein the multi-source optical data comprises an original color image, a black field image, a flat field image, exposure time, camera gain, light source driving current, ambient illuminance, LED pixel gray average value, and corresponding reference exposure time, reference camera gain, reference light source driving current, reference ambient illuminance and reference LED gray average value of each station; Dividing the gain of the current camera by 20 based on 10, and calculating the gain as an index to obtain a current gain amplitude coefficient; The method comprises the steps of multiplying reference exposure time, a reference gain amplitude coefficient and a reference light source driving current, dividing the reference exposure time, the reference gain amplitude coefficient and the reference light source driving current by the sum of a constant one and reference ambient illuminance to obtain a reference energy value, multiplying the current exposure time, the current gain amplitude coefficient and the current light source driving current by the sum of the constant one and current ambient illuminance to obtain a current energy value, calculating the ratio of the reference energy value to the current energy value, calculating the ratio of a reference LED gray average value to the current LED pixel gray average value, respectively taking natural logarithms of the two ratios, adding the natural logarithms, multiplying the natural logarithms by the natural logarithms to obtain a brightness consistency value, and executing natural exponential operation on the brightness consistency value to obtain a brightness normalization value.
  3. 3. The method for analyzing LED color images based on multi-station acquisition according to claim 1, wherein the specific process of forming the corrected input image with consistent illumination is: The preprocessing flow is executed for the original color image of each station: Calculating a brightness normalization value of each station as a brightness compensation gain, and multiplying each pixel value of the original color image by the brightness normalization value to obtain a brightness normalization image; subtracting the brightness normalized image pixel by utilizing a black field image of the station, eliminating camera fixed mode noise and dark current bias, and obtaining a dark field correction image; dividing the dark field correction image by the difference value of the flat field image and the black field image pixel by using the flat field image of the station, eliminating the uneven illumination distribution and the optical path difference of the lens to obtain the correction image, and writing the correction image into a multi-station image analysis database.
  4. 4. The method for analyzing LED color images based on multi-station acquisition according to claim 1, wherein the specific process of extracting color features of standard color areas in standard LED color plate images, fitting to obtain a station color mapping matrix, and performing color transformation on corrected input images pixel by pixel to obtain standardized corrected images under unified reference chromaticity space is as follows: Obtaining standard LED color board images of each station, executing a preprocessing flow to obtain color board correction images, extracting regional average red, green and blue component values in each standard color area in the color board correction images, and constructing observation color data of each station; comparing the observed color data of each station with the corresponding reference chromaticity vector in the reference chromaticity library, obtaining a color calibration mapping matrix of each station through least square fitting, and mapping the observed color of the station to a reference chromaticity space based on the color calibration mapping matrix; The method comprises the steps of inputting red, green and blue components of each pixel into a corresponding color calibration mapping matrix for each station to generate standardized color data of each pixel to obtain a standardized correction image, converting the standardized correction image from an RGB color space to a reference chromaticity space, extracting a brightness value and two chromaticity components of each pixel, and writing the observation color data, the standardized correction image, the brightness value of the pixel and the two chromaticity components of each station into a multi-station image analysis database.
  5. 5. The method for analyzing LED color images based on multi-station acquisition according to claim 1, wherein the specific process of positioning the LED light emitting area in the standardized correction image, suppressing noise and local flicker by effective pixel screening, performing robust aggregation on the chrominance components of the effective pixels, and generating the LED area robust chrominance vector capable of resisting illumination and noise disturbance is as follows: Based on the standardized correction images of each station, performing primary screening on the images according to brightness threshold values, and extracting all LED luminous areas by adopting a connected domain analysis method; Calculating the median of all the brightness values in the area, calculating the absolute brightness deviation of the brightness values of each pixel and the median, and obtaining the median absolute deviation; carrying out validity screening on each pixel in the region to obtain a pixel robust value, wherein when the absolute brightness deviation is smaller than or equal to the median absolute deviation, the pixel robust value is assigned to be 1, otherwise, the pixel robust value is assigned to be 0; the method comprises the steps of obtaining an effective pixel set after effectiveness screening, multiplying one chroma component of a pixel by a corresponding pixel steady value for each effective pixel, adding all products to obtain a weighted sum of one chroma component, multiplying the other chroma component of the pixel by the corresponding pixel steady value and summing to obtain the weighted sum of the other chroma component, dividing the weighted sum by the sum of all the pixel steady values to obtain steady average values of two chroma components of an LED luminous area, and forming a steady chroma vector of the LED luminous area by the two steady average values.
  6. 6. The method for analyzing LED color images based on multi-station acquisition according to claim 1, wherein the specific process of completing the indication labeling of the target color class is as follows: The method comprises the steps of reading a reference chromaticity library, simultaneously calculating a steady chromaticity vector of each LED light-emitting area in each station, comparing the steady chromaticity vector with the reference chromaticity vector one by one, calculating a color difference value, selecting a reference chromaticity vector with the smallest color difference value as a matching reference color of the LED light-emitting area, marking the matching reference color number as a target color label of the LED light-emitting area, taking the minimum color difference value as a color difference characteristic value of the LED light-emitting area, and writing the steady chromaticity vector, the target color label and the color difference characteristic value of each LED light-emitting area in each station into a multi-station image analysis database.
  7. 7. The method for analyzing LED color images based on multi-station acquisition according to claim 1, wherein the specific process for comprehensively analyzing the robust chromaticity vector of each station according to the color category and evaluating the long-term color drift index of the station is as follows: For each standard color type, summarizing the steady chromaticity vectors of the LED luminous areas of the same color type in each station based on the target color label of the LED luminous areas, and calculating an average value as a color characteristic average value vector of the station on the color type; the method comprises the steps of calculating an average value as a system-level reference chromaticity vector of a color class based on color feature average value vectors of all stations on the same color class, subtracting the corresponding system-level reference chromaticity vector from the color feature average value vector of the stations based on standard color class number to obtain a difference vector, calculating a two-norm square value of the difference vector to obtain a color offset, calculating the color offset of each color class, accumulating and averaging to obtain a long-term color drift evaluation value of the stations.
  8. 8. The method for analyzing LED color image based on multi-station acquisition according to claim 1, wherein the specific process of determining the compensation intensity according to the long-term color drift index and applying the compensation intensity to the standardized correction image pixel by pixel to perform the self-closed loop color correction is as follows: calculating a long-term color drift evaluation value of each station And comparing with the multistage drift thresholds T1 and T2; When (when) When the color consistency of the station is less than or equal to T1, judging that the color consistency of the station meets the requirement, and not executing color compensation on the current station; when T1< > When the color of the station is less than or equal to T2, judging that the light deviation exists, constructing a chromaticity compensation vector according to the corresponding difference vector, calculating a light compensation gain coefficient according to the relative positions of the long-term color drift evaluation value of the station and T1 and T2, and implementing chromaticity compensation on the standardized correction image of the current station pixel by pixel, wherein the product of the light compensation gain coefficient and the chromaticity compensation vector is respectively added to two chromaticity components of each pixel; When (when) When T2 is carried out, the color of the station is judged to have significant offset, a chromaticity compensation vector is constructed based on the difference vector, and a severe compensation gain coefficient is determined according to the ratio of the long-term color drift evaluation value to T2; Writing the long-term color drift evaluation value, the color drift state and the compensated image of each station into a multi-station image analysis database, re-executing the calculation of the steady chromaticity vector of the LED luminous area, the matching of the color types and the updating of the long-term color drift evaluation value on the compensated image, re-calculating the compensation gain coefficient according to the latest long-term color drift evaluation value if the new long-term color drift evaluation value is still in an offset state, repeating the color degree compensation, and stopping the compensation when the long-term color drift evaluation value meets the requirement or reaches the maximum compensation iteration number.
  9. 9. The method for analyzing LED color images based on multi-station acquisition according to claim 1, wherein the specific process of carrying out structured feedback and dynamic versioning management of compensation parameters on the monitoring data of each station to realize standardized output of detection results and self-updating closed loop of parameters is as follows: For each LED luminous area in each station, station numbers, LED numbers, steady chromaticity vectors of the LED luminous areas, target color labels, color difference characteristic values and corresponding long-term color drift evaluation values recorded in a multi-station image analysis database are read, and are packaged according to the station sequence and returned to an upper computer through a communication interface; and simultaneously, carrying out versioning management on the color calibration mapping matrix, the compensation gain coefficient and the system-level reference chromaticity vector of the current station according to the compensation execution state, and writing the latest parameters into a multi-station image analysis database.
  10. 10. LED color image analysis system based on multistation gathers, its characterized in that includes: the multi-station data acquisition processing module is used for acquiring multi-source optical data of each station and standard LED color plate images, weakening illumination difference of different stations through a brightness normalization method and forming corrected input images with consistent illumination; The color calibration and standardization module is used for extracting color characteristics of a standard color area from the standard LED color plate image, fitting to obtain a station color mapping matrix, and performing color conversion on the corrected input image pixel by pixel to obtain a standardized corrected image under a unified reference chromaticity space; the LED region robust chroma module is used for positioning an LED luminous region in the standardized correction image, filtering and suppressing noise and local flicker through effective pixels, performing robust aggregation on chroma components of the effective pixels, generating an LED region robust chroma vector capable of resisting illumination and noise disturbance, and finishing indication labeling of target color categories; the cross-station color consistency compensation module is used for comprehensively analyzing the steady chromaticity vector of each station according to the color category and evaluating the long-term color drift index of the station, determining the compensation intensity according to the long-term color drift index, and applying the compensation intensity pixel by pixel on the standardized correction image to carry out self-closing ring color correction; and the result feedback and parameter management module is used for carrying out the structured feedback and dynamic version management of the compensation parameters on the monitoring data of each station, and realizing the standardized output of the detection result and the self-updating closed loop of the parameters.

Description

LED color image analysis method and system based on multi-station acquisition Technical Field The invention relates to the technical field of LED image color analysis, in particular to an LED color image analysis method and system based on multi-station acquisition. Background The LED is used as a core device of a modern photoelectric display and indication system, is widely applied to the fields of consumer electronics, automobile industry, industrial equipment, man-machine interaction interfaces, intelligent manufacturing and the like, and the color stability and the cross-equipment consistency of the LED are directly related to the quality of products, the accuracy of user identification and the reliability of multi-station collaborative production. In the practical production environment of multiple stations, multiple batches and long period, the luminous color of the LED is influenced by factors such as production batch difference, device aging, environmental condition change, station optical structure difference and the like, so that standardized acquisition, standardized expression and consistency management of the color state of the LED become basic links in the industrial LED application scene, and are also key preconditions for ensuring visual consistency and quality controllability. For example, the invention patent with the bulletin number of CN116309887A discloses a method for testing the luminous color of an LED, which is characterized in that an image which is irradiated on a test background board after the LED to be tested is lightened is acquired by a detection camera, the color of the image in an analysis area corresponding to the LED to be tested represents the color of the LED to be tested, the color is analyzed by color analysis software, and then the color is compared with the standard color of the LED to be tested, so that whether the luminous color of the LED to be tested meets the factory specification is judged, the whole process is not needed to be manually judged by operators, errors are not easy to occur, the implementation efficiency is higher, meanwhile, the whole process can be finished only by using a common detection camera, a test background board and color analysis software, the system cost is extremely low, and the method is applicable to various LED manufacturers. For example, the invention patent with publication number CN114612581B discloses a method for detecting color and a detection light source thereof, which relate to the field of color detection, and include that a red light source, a green light source and a blue light source are adopted to respectively shine an object, and a black-and-white red channel photograph, a black-and-white green channel photograph and a black-and-white blue channel photograph are respectively shot to obtain, then three images are synthesized into a color image by using an algorithm. However, when the system adopts a multi-camera and multi-station collaborative detection structure, the spectrum response, white balance, lens distortion and station illumination angle difference of different cameras enable LEDs with the same color to present different RGB or HSV distribution at different stations. Along with environmental changes, equipment aging and light source attenuation, the inconsistency is continuously accumulated, so that color gamut judging intervals are overlapped, and certain stations have cross-station color bleaching phenomena such as red to orange, green to yellow and the like, and the overall consistency is seriously affected. Therefore, in view of the above problems, there is a need for a method and a system for analyzing LED color images based on multi-station acquisition. Disclosure of Invention Technical problem to be solved Aiming at the defects of the prior art, the invention provides an LED color image analysis method and system based on multi-station acquisition, which solve the problem of cross-station color drift inconsistency caused by the influence of optical deviation and environmental change on the color of an LED under multi-station and multi-camera conditions. Technical proposal The LED color image analysis method based on multi-station acquisition comprises the following steps of S1, acquiring multi-source optical data and standard LED color plate images of all stations, weakening illumination difference of different stations through a brightness normalization method to form a corrected input image with consistent illumination, S2, extracting color characteristics of a standard color area in the standard LED color plate images, fitting to obtain a station color mapping matrix, performing color transformation on the corrected input image pixel by pixel to obtain a standardized corrected image under a unified reference chromaticity space, S3, positioning an LED luminous area in the standardized corrected image, screening and inhibiting noise and local flicker through effective pixels, performing robust aggreg