Research On Visualization And Feature Extraction Of Internal Pests In Single Wheat Kernel Based On Micro-CT

There are about 20 million tons of grain lost during warehousing in China every year,of which millions of tons are lost due to stored grain pests.The concealment of early infection of stored grain pests has great harm to the storage,transportation and processing of grain,but the existing detection technology of stored grain pests could not effectively detect the early infection of grain pests.Micro-CT system could obtain three-dimensional information with micron resolution in grain,which is a potential method to realize automatic detection of insect pests in grain.Under the conditions of temperature 30?and humidity 70%,the 3rd,9th,17 th,22nd and 28 th days after mixing Sitophilus oryzae Linne and perfect wheat grain were determined,which were egg stage,young larva stage,old larva stage,pupa stage and adult stage respectively.Micro-CT imaging system was used to collect projection data of granules and infectious granules of different ages.The original projection data are analyzed and the wheat grain projection image is segmented according to the gray value change in the column direction of the projection image.According to the gray scale interval of wheat grain,the gray value of projection image is transformed from [8000,29000] to [0,65535] by linear stretching method.Otsu method was used to automatic segmentation of reconstructed slices.The segmentation image was processed by edge suppression to remove the sample tube wall.The image noise was removed by marking the connected area,and the slice data of wheat grains were obtained.The volume data were constructed from wheat slice images,and the volume data were rendered by marching cubes method and ray casting method respectively.Marching cubes method has the advantages of fast rendering speed and smooth interface,mainly depicting the change of infection from wheat grain surface.The ray casting method is slow in rendering and slow in interfacial interaction.It mainly depicts the change of infection from the inside of wheat grains.Two-dimensional slice data of wheat grains at different ages were analyzed.The results showed that the contour of Sitophilus oryzae Linne changed from single oval to multiple oval combinations,the gap between Sitophilus oryzae Linne and wormhole increased gradually,and the wormhole expanded from epidermis erosion to groin of wheat grains to near epidermis.According to the three-dimensional visualization data of wheat grains,the shape of Sitophilus oryzae Linne undergoes four processes: rod,sphere,ellipsoid and complete body.The length of Sitophilus oryzae Linne changes from 0.37-0.5mm to 2.7-4.3mm,and the width from 0.26-0.3mm to 0.97-1.3mm.The volume of Sitophilus oryzae Linne reaches its maximum at pupa stage.The wormhole extends gradually from epidermis to the center of wheat grains and eventually penetrates through the whole grain.By analyzing the differences of three-dimensional visualization images between infected particles and perfect particles at different ages.Gray value feature,histogram probability distribution feature,salient texture feature and morphological feature of wormhole were extracted to describe the change of wheat grain infection.The 37-dimensional primitive features of energy,sphericity and entropy are constructed.Ant colony algorithm was used to select 17 features with fitness of 91.84823.Support vector machine was used to discriminate 120 wheat grains.The accuracy rate of discriminating whether wheat grains were infected by rice weevil was 100%,and the correct rate of identifying age was 84%.Therefore,micro-CT imaging technology is feasible as a method to detect early infection of wheat grains.