Research On Image Segmentation Algorithm Of Orange Surface Defects

Segmentation and identification of citrus surface defects is an important link in the process of acquisition and sales of fruit,which is of great significance to improve the quality of fruits and enhance the added value of fruits.This research choose eight common citrus defects insects,decay,anthrax,wounds,wind trauma,thrips,ulcers as the research object.The classic algorithm,the GAC model algorithm and the LBF model algorithm are used to segment the experiment object.The research contents are as follows:Segmentation experiments have been implemented to citrus defect images by three kinds of the traditional detection method,including threshold method,region method and edge detection operator.The research selected Two-dimensional information entropy method of the threshold method,Roberts operator,Sobel operator,Canny operator,Prewitt operator,LoG operator of the edge detection,the regional seed growth method of the Regional law as the experimental method.Record and analyze the advantages and disadvantages of each algorithm and its application.By analyzing the experimental results,observe the improved ability of GAC model to recognize eight kinds of defects on citrus surface,and then verify the feasibility of improving GAC model.In this paper,the traditional GAC model is studied in-depth theoretical,and it would be improved.An improved GAC model was used to test citrus defect images and the parameters that affected the segmentation results would be recorded.The traditional LBF(Local Binary Fitting)algorithm and the improved LBF(Local Binary Fitting)algorithm were used for comparison experiments.By comparing the segmentation time of the two algorithms,segmentation results,the scope of the algorithm to verify the superiority of the improved LBF(Local Binary Fitting)algorithm.By the way of comparing and analyzing the improved GAC model algorithm and improving the segmentation result of LBF algorithm,The conclusion is that the improved GAC model has better effect on segmentation of sag defects,but its anti-noise performance is poor.The improved LBF model has the best comprehensive segmentation effect and fast operation speed,but it has higher requirements for early stage parameter adjustment.The model algorithm applied to the image segmentation of the citrus defects was made into a software experiment system in the end.