Study On Identification Method Of Delinted Cottonseeds Varieties Based On Hyperspectral Image Technology

The quality of seeds is related to the development of the entire agricultural production,in which the identification of seed varieties is an important means to ensure the quality of seeds.The area of cotton planting in China is relatively large,hybrids develop rapidly.The variety emerges endlessly,and the variety disorder,miscellaneous and other phenomena are common..Because of the traditional detection methods,there are many shortcomings such as long cycle,heavy workload and cumbersome process.Therefore,this paper combines the actual requirements of agricultural production and introduces hyperspectral imagery techniques into the classification and detection of delinted cottonseed varieties,in 710,Xinluzao 41,Xinluzao 50,Xinluzao 57,and Xinluzao 62 as the object of study.By taking full advantage of the“spectrum and image integration” of hyperspectral images and combining a series of stoichiometric methods,a convenient,efficient,and quick method for non-destructive identification of delinted cottonseed varieties was developed.The main findings are as follows:(1)Fusing spectrum and image information of hyperspectral image to identify the species of delinted cottonseeds.The SG smoothing and normalization methods were used to preprocess the spectral data of cottonseeds.Using active contour model to extract the length,width,area,circularity and other 12 morphological parameters of the cottonseeds.SPA-GA-PLS projection analysis were used to select the spectral and image feature variables of cottonseeds,the selected seven spectral feature variables and five shape feature variables were fused.Using PLS-DA,SIMCA,kNN,PCA-LDA and PCA-QDA methods to establish the cottonseeds classification model based on the fusion of spectral and shape information.The results are as follows: PLS-DA model has the best prediction effect.The recognition accuracy of the fusion model raise from 93% to 96%,which indicates that the spectral and image fusion information of hyperspectral image can effectively improve the recognition accuracy of the model.(2)Using hyperspectral image information to explore the effect of seed aging on the classification of delinted cottonseeds.The artificial accelerated aging treatment three types of delinted cottonseeds,Xinluzao50,Xinluzao 57,Xinluzao 62,was divided into three grades: unaged treatment,aging 24 h,and aging48 h.Using the SPA combined with partial least-squares projection algorithm,combining selected characteristic wavelength with PLS-DA,SIMCA,kNN,PCA-LDA and PCA-QDA to establish the classification model.It reflects the effects of seed aging on the internal and varietal differences of varieties.The results show that the average recognition accuracy of the six types of PLS-DA models is 96%.Hyperspectral image technology can reflect the influence of seed aging on difference of cottonseeds to a certain degree.Thus,the characteristic changes of the delinted cottonseeds during storage aging areexplored,at the same time,it provides technical support for the intelligent monitoring of the delinted cottonseeds storage process.(3)The visual identification of mixed delinted cottonseeds based on hyperspectral imaging technique.The hyperspectral mask image at the characteristic wavelength of the mixed delinted cottonseeds was used as input variables and imported into the previously constructed PLS-DA classification model,and all pixel points on the cottonseed samples were predicted.Combined with the image processing technology,a pseudo-color visualization map of mixed cottonseeds was obtained.Finally,the visual identification of mixed cottonseeds was realized.