Maize variety identification method based on deep-cascaded forest and hyperspectral image

Abstract

The invention provides a maize variety identification method based on deep-cascaded forest and a hyperspectral image. The method comprises the following steps: pretreating all maize image data to obtain full-waveband image data selecting the maize image data in the effective waveband; and taking the maize image data in the effective waveband as a feature set, and performing model classification byadopting a deep-cascaded forest model. The maize classification is performed by sufficiently utilizing the hyperspectral technology and the deep-cascaded forest model, compared with the ordinary machine learning algorithm, the maize variety identification method disclosed by the invention has higher identification precision.

Description

Technical field [0001] The invention relates to a method for identifying corn varieties, in particular to a method for identifying corn varieties based on deep cascade forest and hyperspectral images. Background technique [0002] Corn is one of the important food crops in my country, accounting for more than 30% of the total grain output, and the annual seed production exceeds 1 billion tons. With the diversification of corn demand, my country's corn varieties are also increasing year by year. In the breeding process, some illegal traders pretend to be shoddy and use inferior corn varieties to pretend to be high-quality corn varieties, resulting in a decline in corn production and bringing huge hidden dangers to national food production and agricultural safety. Therefore, how to accurately and efficiently identify corn seed varieties without damage It is of great significance. [0003] The current methods of identifying corn seeds mainly include manual inspection, physical and c...

AI Technical Summary

Problems solved by technology

Manual detection mainly relies on the naked eye to make judgments based on the characteristics of corn size, shape, and color. This method is greatly affected by human subjective factors. Not only is the detection accuracy low, the speed is slow, and the labor intensity is high, it is difficult to form a unified standard.

Physical and chemical detection is mainly based on the biochemical characteristics of different varieties of corn to achieve identification. This type of method has a high accuracy rate, but it can only be detected by sampling, which is damaging and the process is complicated. The operation is cumbersome and it is difficult to meet the market demand.

Computer vision detection is mainly based on the near-infrared spectral information of seeds and the morphological feature information of visible light, and the application of image pattern recognition technology to establish a variety discrimination model has the advantages of instant, efficient, non-destructive and accurate; but this type of method extracts less feature information , and most of them are based on the evaluation of external features, and the features that characterize their internal components cannot be obtained, and the reliability of the test results is affected; in addition, with the increase of seed variety data, the cross phenomenon of seed features is serious, which leads to the deterioration of the separability of the model

Images