The Study On Microscopic Image Processing Method Of Wheat

The distribution and use of wheat is widely,which is an important grain crop that it’squality has aroused widely concern. With the continuous development of science andtechnology, people have put forward higher requirement for the detection of wheat quality.While the he traditional detection methods existence some defect that are difficult toovercome, information technology, especially image processing improved the development ofdetection technology. At present, the application of digital image processing technology inmicroscopic images of wheat is still few, in older to provide technical support for furtheranalyze the microstructure of wheat and provide the theoretical basis for the establish anobjective, accurate and rapid wheat grain quality detection, microscopic image processingmethod of wheat was researched in this paper.SEM image may present edge blur and low contrast phenomenon for the focusingdeviation or other reasons. Aiming at this problem, a low-contrast microscopic imageenhancement method was proposed. It used Sobel gradient transform to highlight the edgeinformation and emphasized the details of image by Laplace of Gaussian filter, then enhancedthe relative low frequency image with improved contrast limited adaptive histogramequalization, the ultimate enhanced image was obtained by the weighted sum of the previousthree processed images. Experimental results show that the method can enhance thelow-contrast SEM image effectively which is propitious to the subsequent image analysis andfeature extraction.The size and size distribution of starch granules are the important quality characteristicsof starch. A method of wheat starch granules edge detection was proposed in this paper. First,extracted the image edge gradient information by using local maximum difference methodand weighted it to the original image to enhance the edge of starch granules. Then smoothedthe image by morphological reconstruction operation which have good performance inpreserving edge features. finally, use Canny operator to detect the wheat starch granules edge.Experimental results show that the method can extract the edge exactly, which helpsresearchers for the further quantitative analysis of microstructure of wheat and completing themeasurement of starch granules size parameters. Traditional machine vision-based classification of wheat cultivar extract the appearancecharacteristic such as shape, color or texture. As different wheat cultivars showed differentgrain microstructure with the form of gray, a classification of wheat cultivar base onmicroscopic image was proposed. It first extracted gray level co-occurrence matrix (GLCM)texture features of microscopic images, then the best features or their combination were beselected as the input classification of BP neural network, finally, network classifiers were usedto classify the five kinds of wheat cultivars, that is8802-1,YangMai11, YanNong19-1,XiNong979and AiZao. Experimental results show that the overall classification accuracy isabove90%, The method is simple and it presents a new way for the classification of wheatcultivar.