Study On Nondestructive Testing Method Of Potato External Defect In Natural Light Environment

Potato as the fourth largest staple food in China,the detection of its external defects directly affect the industrialization and deep processing of potato.At present,most of the domestic detection are based on specific circumstances,which does not meet the requirements of real environmental detection,therefore,this paper proposed in the open natural light environment on the detection of potato defects,including green skin,germination,disease and normal,which is of great significance for deep processing and commercialization of potatoes.In this paper,we first study the segmentation of potato image in natural light environment,and propose a new method to segment the potato target area,First,the background color of the potato is eliminated by setting the threshold of R-B and G-B components,then through the saturation grayscale processing and Otsu method for segmentation,Last with morphological filtering to remove holes and other noise,and finally the potato area were segmented.Then,a three-layer convolution neural network model is designed to classify,with the potato sample target segmentation pre-processing,then input to the three-layer convolution neural network to learn and test,and finally the normal potato,green potato,germination and potted potato samples were classified.Finally,the corresponding detection method is proposed for the defect area.Aiming at the green skin region,a new method based on RGB,HSV and LAB multi-color model is proposed,which mainly divides the color component difference R-G of the RGB model,the Hue of the HSV model and the Chroma A of the LAB model.And the method of fusion of the characteristic parameters is carried out.Finally,a new method for detecting the gray variance of Laplace operator is proposed for the germination and lesion area.The average of the small area is extracted by Laplace sharpening variance and the average variance threshold of the whole potato target area were compared,and finally the germination and lesion areas were successfully located.