Research On Nondestructive Detection Of Pecans Rancidity Based On Magnetic Resonance Imaging

Pecan,which is produced in the United States and Mexico,is one of the top ten nuts in the world,.Because of its good taste and rich nutrition,it is loved by people.In the process of production and processing of Pecan,it may be oxidized,which will lead to the phenomenon of Pecan rancidity,and seriously affect the taste and sales of Pecan.In addition,what is more important is that taking rancid nuts will do harm to human body in many ways.Therefore,it is of great significance to detect the existence and degree of pecan rancidity.The research on Nondestructive Detection of pecans is conducted based on nuclear magnetic resonance imaging technology to achieve the goal of detecting the existence and degree of pecan nut rancidity more accurately by using its advantages of fast imaging,high tissue resolution and non-invasive.The concrete contents are as follows:(1)In order to solve the problem of uneven gray level of Pecan's NMR image and the difficulty of using commonly used segmentation methods to accurately segment the nuts in the image,an adaptive level set of segmentation algorithm based on improved edge indicator function is proposed in this dissertation.In this algorithm,the local gray fitting values,which inside and outside the curve,are added to the edge indicator function,and by combining the gradient information after image convolution,the disadvantage of the curve evolving into a false boundary is avoided.On the one hand,variance information and image information function are added to the area term to make the direction and velocity of curve change adaptively.On the other hand,Local Gray Fitting Values inside and outside Curves and Gradient Information after Image Convolution are draw into the edge indicator function to avoid the disadvantage that the curve evolution falls into pseudo-boundary.In addition,the distance regularization term of double well potential function is added to the model to avoid the periodic initialization of level set function during the curve evolution process,meanwhile,the computational complexity is reduced while the evolution speed is improved.Finally,the precise outer contour segmentation of pecan nuts is realized,which promotes the subsequent processing.(2)In order to solve the problem of noise in image caused by current or magnetic field in NMR imaging technology,which has effect on the performance of feature recognition,an improved anti-noise Local Binary Pattern(LBP)descriptor is proposed.In this algorithm,the fuzziness mechanism that using the Dixon criterion in statistics to find the outliers of each 3*3 neighborhood in the image which is the fuzziness given is adopted.New gray value of the ambiguous pixel,of which,whether the mode is 0 or 1 is determined by the consistency of neighborhood pixels,is assigned by calculating the gray mean of four symmetrical left-right pixels in the annular neighborhood of the point,which can achieve the resistance of noise interference.Finally,the improved anti-noise LBP descriptor proposed has a great performance on feature extraction,classification and recognition of pecans based on MRI images.(3)To solve the problem that it is impossible to determine the acidity of pecan only by means of a single MRI slice image of pecan,the comprehensive consideration is given in this dissertation.While each pecan can obtain 5 sequential MRI images,whether the rancidity exists in pecans is determined by the situation that there is at least one slice judged to be rancid of the five MRI slices,otherwise the pecan is considered to be normal.On this basis,a self-defined formula is used to calculate the weighted sum of suspected rancidity area percentages in each of the five MRI slices of pecan.Finally,four grades of pecan rancidity are classified: normal,mild,moderate and severe.