Research On The Orientation And Classification Of Apples Based On Machine Vision

At present,due to the lack of the automatic detection technology,most of the domestic agricultural producers chose to classify agricultural products by manual detection.However,it is not only difficult to improve production efficiency,but also difficult to unified the evaluation criteria of agricultural products.The use of machine vision can realize the nondestructive testing of the fruit quality.Meanwhile,it can effectively unite the quality of the fruit and improve the production efficiency and market competitiveness of the enterprise.This paper takes Red Fuji Apple as the research object,and focuses on the key technologies of apple classification system based on vision,mainly involves the design of apple orientation mechanism and apple classification algorithm.Specific research contents are as follows:1.Through the analysis of object,apple grading standards are identified.The standard includes both the specification level and the quality level.For the classification efficiency and classification accuracy,the automatic grading production line with orientation mechanism and multi-camera is selected.Also,the specific mechanical structure design is completed.2.The mechanical model of the orientation mechanism for apples is established,and the collision parameters are determined.Through the ADAMS kinematics simulation analysis,it is verified that the orientation mechanism can realize the automatic orientation of different pose and different specifications,and the apple will not be damaged during the orientation process.The influence of the mechanical structure parameters on the orientation performance is discussed by the shortest directional time orthogonal test,and the optimal scheme is determined.3.According to the actual requirements of the visual inspection system,the hardware is selected and industrial camera calibration is completed.An improved color model of HSI is proposed which constructs a new color index by using the three component of the color model.And the Otsu algorithm is used to segment the new color index,realizing the background segmentation of the apple images.The undistortion of the apple image is realized by combining the undistortion algorithm flow.In order to suppress the difference of the surface brightness of apple and improve the segmentation efficiency of suspected defects,a fast segmentation method for apple defect based on edge brightness correction is proposed.4.In order to improve the efficiency of finding parameters of SVM,a chaos-based particle swarm optimization algorithm is proposed in this paper.The test result shows that the algorithm can avoid falling into local extreme points and the calculation accuracy could be improved.5.The algorithm involves four indicators,which are the size and defect size,fruit shape,and color.The extraction process is as follows.Each pixel in the X and Y direction of the actual representation of the length is calibrated,and the correct rate of apple size classification is 92.9%.The Zernike moment is used to detect the shape of the apple and the dimension of the first six Zernike moments are reduced.The first three principal components are used as the fruit characteristic parameters,and the rate of classification accuracy can reach to 93.1%.The mean of RGB and HSI color components is used to construct color feature parameters,and three parameters is utilized to describe the color characteristics after reducing the dimension,and the rate of classification accuracy gets 100.0%.The double tree complex wavelet is used for apple fruit stalk and defect recognition,and the correct rate of defect segmentation is 84.6%.6.Based on the Matlab software,the GUI design of the grading program is combined to synthesize the apple fruit surface shape,color,defect size and size information.There are three major parts,which are the landing system,online test system,off-line test system.The experimental results show that the comprehensive accuracy rate of the grading system is 86.8%.