Based On Image Processing Of The Appearance Quality Inspection And Grading Of Wheat

Food is an indispensable part of people’s daily lives, is China’s second-largest food crop, accounting for an important position in the national economy; establish a system, scientific and effective food quality inspection system is of great significance. The purpose of this study is to explore and analyze the new method based on image processing technology, the detection and classification of wheat quality through improved feature extraction algorithm to find the characteristic parameters of varieties to identify the contribution rate, establishing the optimal model able to correctly identify wheat varieties provide technical support for the automatic detection of a breed of wheat classification.This article selects the (L8998、Neixiangl88、9023、Youzhan1、Yumai47、Zhoumai12), six un-classified wheat kinds as the research object, first of all, the color of the wheat grain were extracted respectively characteristic parameters (red, green, blue, hue, saturation and brightness of the mean and variance); Morphological parameters (rectangular, circular, perimeter, area, elongation, ovality, the elliptical eccentricity, equivalent diameter) and texture parameters (energy, entropy, contrast, moment of inertia and local stationarity) etc. A total of19eigenvalues. Then the6classes using BP neural network for classification of wheat:1) morphological parameters and color parameters as the classification of input output;2) will form, color and texture characteristic parameters as the input of the classification results together. Finally, according to the classification results of BP neural network is optimized, and use the MES (target error) and MIV-BP algorithm characteristics of input parameters for the selection, analyzes the relations between various input parameters and the wheat grain weight and the impact on the classification of output.Experimental results show that the optimized BP neural network classification recognition rate respectively for6classes of wheat:L8998wheat recognition rate is88.5%, NeiXiang188wheat recognition rate is93%, and9023wheat exhibition of recognition rate is92%, YouZhanl wheat recognition rate is87.5%, yumai47wheat recognition rate is90%, Zhoumai12wheat class recognition rate is87.25%, the sample recognition accuracy rate reached93.5%than before the network optimization of sample identification accuracy increased by82.05%; After a variable filter, therefore, the output of the neural network classification accuracy obviously raised, so as to achieve the aim of network optimization.