Click on “Download PDF” for the PDF version or on the title for the HTML version.

If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.

Recognition method of origin Gastrodia elata based on machine vision

Published by the American Society of Agricultural and Biological Engineers, St. Joseph, Michigan

Citation:  2021 ASABE Annual International Virtual Meeting  2100099.(doi:10.13031/aim.202100099)
Authors:   Zimei Zhang, Zhian Zheng, Lei Gao, Rongyan Wang
Keywords:   Bayesian; features extraction; Gastrodia elata; identification of origin; image recognition; principal component analysis.

Abstract. Gastrodia elata Blume (G. elata) is a precious traditional Chinese medicine, and its quality is crucial to its efficacy. Because of the difference of growth environment and climate, the quality of G. elata is quite different to different places of origin, which leads to changes in medicinal properties. So the identification of places of origin of G. elata becomes an important part of their quality control. However, the traditional sensory recognition method is too subjective and the analysis method of chemical composition which often lags behind is difficult to comprehensively analyze the Chinese herbal medicines. In order to identify the origin of G. elata objectively and efficiently, we proposed a method based on machine vision technology, which realized the featuress extraction and image recognition of G. elata in two typical producing areas of Yunnan and Anhui provinces in China. Firstly, 295 typical images of G. elata in two places were obtained. Secondly, in order to obtain clear images, they were subjected to pre-processing including compression, denoising, opening operation and segmentation. Thirdly, 18 features of the third-order moment components of RGB and HSV color space were extracted respectively and reduced to three dimensions by principal component analysis (PCA) to improve the recognition speed. Lastly, by establishing a Bayesian discriminant model for sample training and testing, the final overall recognition rate reached 90.5%, which effectively realized the recognition of G. elata images in the two places. The results indicated that the PCA algorithm and the Bayesian discriminant model based on color features can effectively realize the image features extraction and classification of G. elata from Yunnan and Anhui which were two typical producing areas, providing a basis for the evaluation of G. elata quality.

(Download PDF)    (Export to EndNotes)