Research On Leaves Image Color Correction And Its Application Based On The Reflection Model

Green is mainly presented by the changes in photosynthetic pigment, and varies with the photosynthetic pigment content distribution, which often also shows various physiological characteristics of trees. For machine vision recognition and color difference analysis, it needs to make sure that recognition is in the same constant illumination conditions so as to keep the color constancy, removing the light effect and keeping color constancy to make the comparable color and choose the appropriate color space to ensure the consistency of the color difference. This paper mainly content is as follows:1、Study the color mechanism and the physiological properties of plant leaves, including leaf color pigment composition, different pigment for physiological role in plants, and the changes of physiological characteristics with leaf color changing.2、According to the Lambertian reflection model, the qualitative analysis of the existing without the supervision of the basic theory of color constancy, including White Patch algorithm, Grey world theory, Grey Edge frame and Von Kries model, uses the SFU database image in the Grey Edge framework and determinates the parameters; Quantitative methods analyze the color test plate color difference between the real color and image detection plate color, and determine the validity of the algorithm.3、Based on the Lambertian reflection model in the airspace and frequency domain, Retinex theory is studied. Image enhancement methods Retinex theory, SSR. MSR and MSRC R algorithms are given, and analyze these characteristics and limitations; In addition, based on homomorphic filtering model, we use the advantages of uniform Lab color space, light weight and chrominance component separation, only deal with illuminance component and obtain the good effect.4. Analyze the characteristics of different color spaces, including RGB, XYZ, HSI, LUV and Lab. According to effectively and obvious color features of the leaves in the image, find ways to eliminate interference and increase color features:1) eliminate the influence of illumination for color features;2) determine the color similarity measure reflecting off color;3) determine the optimal threshold. In this paper, using the H component in HSI space, the color consistency of human eye vision system is used to determine the threshold value for image segmentation; in Lab uniform color space, conduct the chromaticity component clustering analysis based on Euclidean distance, and compare with the existing method for image segmentation using color features.