Study On Methods Of Detecting Friuts Based On X-Ray Images

The traditional detecting methods on fruits' internal quality are time-consuming anddestructive for the products. Therefore, the research on rapid, objective and nondestructivedetecting methods for agricultural products quality has becoming the hotspot. This paperapplies the image processing, genetic algorithm and artificial neural network technique to thenondestructive detecting for agricultural products quality on the basis of the given chestnuts'X-ray images. The main points are followed:(1) This paper preprocesses the given chestnuts' X-ray images and chooses averagefilter, Wiener filter, median filter, Butterworth filters and exponent low pass filter to filter theimages. The results showed that the3×3median filter is the best for de-noising.(2)This paper proposes a method based on the combination of the histogram concavityand the sum of the neighborhood differences. In addition, the comparison was done when theproposed method and2D histogram methods were applied in segmentation on X-ray imagesof fruits. The results showed that the proposed method's inaccurate area is less than2.1%andthe max inaccurate area is23.7%of two-dimensional gray histogram's inaccurate area.(3) This paper choesed24features for genetic algorithm to optimization grouping on thebasis of fuits' X-ray images. This paper determined the encoded mode, the size of initialgroup which is300, number of generations whicih is10000and chosed fitness function,selection operator, crossover operator, mutation operator. The result of genetic algorithmshowed that the best group of features is consistence, correlation, inhomogeneity, grayaverage, mean square deviation of gray, inertia and inverse difference moment.(4)This paper chose BP network whose structure is7-10-2to process the images andused genetic algorithm to optimize the network's initial weight and threshold. The optimizedgroup is the input vector of the network which is trained then. This paper uesd the test set totest the recgnition tatio, and the result showed that the proposed method' recgnition ratio is92.24%。This paper had done the research of detecting the internal quality of fruits and offered anew idea which is critical to improve the level of agriculture products quality's detection．...