Objective  In order to solve the problem of variety confusion and low utilization rate in Chinese chestnut, geometric morphometry was used to digitally analyze the nut morphology of different Chinese chestnut varieties, and the variety identification method was established.

Method A total of 3200 nuts of 80 varieties from different chestnut producing areas in China were used as test materials, and the plane, side and bottom components were photographed with unified parameters. Using Image J software combined with the morphological characteristics of chestnut nuts, 17, 12, and 16 identification marks were selected in different components in a specific order so as to obtain nut morphological coordinate data. Morpho J software was used to perform generalized procruste analysis on the data to eliminate the interference of non-morphological factors such as size and location, and to classify the data by variety. Performing relative warp analysis based on principal components on the data was to obtain the contribution rates of identification marks in different components, so as to extract important identification marks and visualize the morphological differences of nuts. Canonical variate and discriminant analysis were used to illustrate the identification effect of identification marks on different chestnut varieties.

Result (1) Relative distortion analysis showed that the cumulative contribution rate of the top 25 identification points was 87.7%, which had a greater identification effect and was an important identification marks. (2) Principal component analysis showed that the morphological differences of nuts were mainly reflected in the top and shoulder, the plane contour shape of nuts, the thickness of nuts, and the relative size of the hilum. (3) The analysis of typical variables showed that most varieties had overlap and low differentiation when different components were identified separately. When 34 identification points were used for comprehensive identification, there were significant differences among different varieties, and the identification effect was obvious. When 25 important identification marks were used, the effect was similar to that of the former. (4) The discriminant analysis of 80 varieties for different components showed that the accuracy of plane, side and bottom components were 72.5%−100.0%, 71.8%−100.0%, 70.0%−100.0%, respectively. Comprehensively discriminated with 45 identification marks, the correct rates were 95.0%−100.0%. 33 important identification marks were used for identification, and the correct rate was 92.5%−100.0%. The latter two had a higher identification rate and could realize the identification of 80 chestnut varieties.

Conclusion The geometric morphometric analysis based on 25 identification marks could achieve accurate identification of different chestnut varieties, with a correct discrimination rate of 92.5%−100.0%. In this study, a method for chestnut variety identification based on digital analysis of nut morphology is established, which will provide a new basis for the correct application of varieties in production practice.