| At the present stage,navel orange growers in China are mainly small and medium-sized individual households,and the main grading methods are still limited to manual methods.Therefore,a miniaturized,low-cost,high-throughput automatic inspection and grading equipment is needed to improve the post-harvest grading methods of navel oranges in farmers' fields.The aim of this study is to design a multi-channel machine vision-based external quality inspection and grading equipment for navel oranges,with the following main research elements.1.Design of navel orange inspection and grading equipment.It included detection module and grading module.In this paper,a four-channel navel orange inspection and grading device was designed,with the device length,width and height of 6715 mm,1077mm and 1256 mm,respectively,with the potential to be installed on a mobile carrier.The inspection module mainly includes the design of conveying rollers and the design of image acquisition device:(1)A 610.3mm long stepped shaft in series with four rollers conveying navel oranges was used according to the inspection requirements;(2)The selection of machine vision hardware was determined according to the camera field of view and accuracy requirements.Physical tests were conducted and it was found that the rollers drove navel oranges to achieve 120° tumbling at adjacent stations.The grading module used a modular,building block type drop grading mechanism for efficient use of space.2.Visual inspection scheme design and light condition optimization.Light Tool S simulation software was used to analyze the light source box with different tetrahedral structures for uniform illumination.Solid Works software was used for solid 3D modeling of the inspection scene.Among them,the upper half surface of the orange model was used as the receiving surface,and the surface irradiance was used as the characteristic parameter for analysis to optimize the pyramid structure with the best lighting effect under the condition that the luminous surface and the diffuse transmittance plate cooperate.The tilt angle between the pyramid slant and the bottom surface was 45°;and the built vision system was used as the basis for live testing to find that:(1)The mean grayscale values of the 12 navel orange areas under the inspection field of view ranged from 205 to 210,with a variance of 31 to 35,which could reduce the error caused by the difference in grayscale when dealing with different navel orange areas at the same time.(2)The mean grayscale values of the fixed 8 areas fluctuated within 10 under the continuous operation of the LED light source for one month,and there was no obvious pattern between the mean grayscale values and the date.3.The method and study of external quality inspection and grading of navel oranges.For size inspection,the size of 142 navel orange samples was measured by the fitted ellipse method,and the measurement error was ± 0.96 mm.For shape inspection,the area contours of 110 samples were analyzed for near-circularity,and the values of 1.180 and 1.200 were used to distinguish normal-shaped fruits,slightly deformed fruits and severely deformed fruits.For color inspection,the color grade was evaluated by the grayscale mean value of the navel orange area on the Hchannel image.For defect inspection,using The histogram equalization method and Canny algorithm were used to obtain the defect edges,the roundness,structure coefficient and grayscale characteristics of the defect area were used to distinguish the defect types,and the classification accuracy of different defects was above 90%,among which the classification accuracy of sunburn,crack,bruise and puncture wound was 100%.4.Software design of navel orange inspection and grading system.PLC was used as the core to realize the motion control of conveying line,grading line and unloading unit.The computer software system was designed using VS2013+QT5.9,and the image processing algorithm compiled by HALCON was called to distinguish navel oranges into 10 grades with the algorithm time less than 300 ms.331 samples were tested online,and the classification accuracy was 92.7%,and the inspection accuracy of equal external fruit was 100%. |
PDF Full Text Request