| Maize is one of the main grain crops in China,quality detection plays an important role in the harvest,storage and deep processing of it.The water content and the breaking rate serve as the important parameters determining the quality of maize.However,most of the detections for maize moisture content have the disadvantages of low efficiency,limited measuring range and being easily affected by environmental temperature,while the detections of breaking maize kernels are mainly carried out manually,which easily leads to subjective assumption and low efficiency.Therefore,the machine vision technology was introduced in the thesis to explore the new method of detecting maize moisture content and breaking rate by combining with the image processing technology.The main research contents and conclusions of this thesis are as follows:(1)Establishment of complete maize kernel moisture detection model.Specifically,four maize varieties(including Xiangyu 218,Jidan 519,Xianyu 335 and JND516)were chosen as experimental material in this thesis,and digital images of complete maize kernels with different water content of maize varieties were collected.Followingly,based on image processing and analysis,moisture detection models of four varieties were established.The results showed that the absolute errors of moisture content detection in the four moisture detection models were all within 1%,while the relative errors were all within 4%.Furthermore,the feasibility of the method in moisture detection of maize across varieties was analyzed in the thesis through establishing moisture detection model of various varieties using three maize varieties including Xiangyu 218,Jidan 519,and Xianyu 335.Furthermore,the test showed that the absolute error range of testing moisture was-6.73% and-2.72%,respectively,and the relative error range was-60% and-13%,respectively.Therefore,it can be concluded that this method has a good effect and high accuracy in the detection of moisture content among the same varieties,but has a big error in the detection of moisture content across maize varieties.(2)Studying on the detection method of damaged maize kernels.Two methods,namely,discriminant analysis method and successive sweep method for identifying damaged maize kernels were proposed in the thesis.Among them,the former determines whether the maize kernels are damaged by establishing discriminant equation.Statistically,among 736 complete maize kernels,30 were misjudged as damaged ones,with an accuracy rate of 95.92%;In the actual detection process,the maximum of 2 in a plate(184 grains)of maize samples were misjudged,with the accuracy rate as high as 99%.The latter is a method that gradually remove the damaged maize kernels from the original images through threshold selection based on the difference in geometric features and the color features between complete maize kernels and damaged ones.The actual measurement test showed that the maximum of 3 in a plate(184 grains)of maize kernel samples were misjudged,with the accuracy of 98.4%.Therefore,it can be concluded that a good detection effect can be achieved using both two methods above.(3)Establishment of moisture content detection containing damaged maize kernels.Based on the fact that the damaged kernels can affect the accuracy of maize moisture content detection,the errors in image information and weight information were corrected in the thesis,and a new moisture detection model was established to detect the moisture content and the damaged kernels of maize synchronously.The results showed that the absolute error range of the moisture content measurement values of the four new models was within 1%,while the detection errors of the damaged kernels in one plate(184 grains)of maize sample was within 2,with the accuracy rate up to 98%. |
PDF Full Text Request