Pepper External Quality Detection Technology Based On Machine Vision

Agricultural products external quality detection technology has always been a hotresearch field of machine vision problem,with the social progress, the level of science andtechnology continue to improve the the computer level high-speed development, theagricultural testing technology more and more applications in machine vision,,has a broadspace for development and development prospects.Detection techniques for agriculturalresearch already underway for many years abroad,it has a wealth of theoretical basis andpractical experience.Detection technology has only just started in China, there is a great gapbetween domestic and abroad.In pepper quality detection,research is still little at home andabroad,article based on shape and color characteristics of the chili,proceed fromreality,preprocessing the algorithm for chili, chili segmentation algorithm, chili featurecalculation and pepper recognition algorithm and laid a solid foundation for further study.Mainly to the completion of the work of this study have the following:1Image acquisition.Image acquisition of this article is divided into two ways that thelaboratory the camera obscura environmental and complex environment.The study is thepepper image in the the laboratory camera obscura environment,so laboratory environment tocollect good quality chili image is the key to the experiment to success.Laboratory cameraobscura environment using a single color for the background, eliminating the complexity ofthe environmental interference to make the acquisition of images with higher quality.Thisarticle pepper samples were collected image100, Category5chili slices image40.2preprocessing algorithm pepper, image segmentation, contour extraction and peppersize detection algorithm.Since the chili has special color characteristics,chili imagepreprocessing algorithm of this paper use the color threshold.Pepper, after dealing with asingle image background,very benefit for pepper and background separation.Chili aftersegmentation algorithm,get the pepper contour extraction to prepare for the the chili size ofthe eigenvalue calculation.3. Hough forest classification application.This paper use hough forest classifier toidentify the chili object,to establish the corresponding random tree by learning sample chili,do probabilistic polling through the image block of the image to be detected to the positionthe pepper may occur.Then get a vote results,according to the the results generated houghimage pepper location.4. Qualitative analysis of the chili color. Because color composition is obviously,mainlyin the red as the main,the lesion more for dark yellow,how much yellow component to determine the chili is good or bad.Due to time constraints, this article does not pepper colorrange of quantitative statistics.5. To eliminate the peppers handle on the the chili size of interference, pepperoverlapping image design not to consider.This study object-pepper, is an attached handlecrops,the presence of the shank in the process of calculating the size has a great impact on theresults.This paper designed an algorithm to eliminate the pepper handle more accuratecalculation.This study is not just research a single pepper,but also size acquisition multiplepepper.But the chili overlapping situations often happens.Due to limited time, the article doesnot given overlap pepper stylistic reasonable algorithm will not consider in this study.Through this research,we get a certain curvature, Petiole or Terrier crops, researchmethods. There is a lot of significance on the crops further study in the future.The same time,due to the limited time of the study, there is insufficient does not solve a lot of the stylistic andalgorithms to resolve urgent need of future research.