Combined-skewness-based waveband selection method for hyperspectral image of corn seed

Abstract

The invention relates to a combined-skewness-based waveband selection method for a hyperspectral image of a corn seed. According to the technical scheme, the method comprises: step a, a to-be-identified corn seed sample is placed in a hyperspectral image collection system and thus hyperspectral images of the corn seed are collected and obtained; step b, a contour curve of the corn seed is obtained by using threshold segmentation, spectrum mean features and entropy features of the corn seed under the contour are solved and obtained, feature combination is carried out and then the combined feature is used as a feature parameter X of the to-be-identified corn seed; step c, a combined skewness value JS (X) in a full-waveband mode is resolved and obtained by using the feature parameter, waveband selection is carried out by using a sequential backward selection method, and then an optimal wave combination phi is outputted; and step d, a prediction model is established and an evaluation result of the waveband selection method is obtained. According to the invention, waveband selection is carried out on the hyperspectral image of the corn seed, so that waveband selection can be realized on the multi-feature condition. Moreover, the method has advantages of simple operation, high rapidity and effectiveness, and high robustness.

Description

technical field [0001] The invention relates to a method for selecting a band of a hyperspectral image of corn seeds, in particular to a method for selecting an optimal band of a hyperspectral image of corn seeds based on joint skewness. Background technique [0002] Seeds are the most important means of production, and the identification of seed varieties is of great value in reducing seed mixing and ensuring the smooth progress of agricultural production. In order to overcome the disadvantages of long identification time and destructiveness of traditional seed variety identification technology, non-destructive identification methods of seed varieties represented by near-infrared spectral analysis technology, machine vision technology, and hyperspectral image technology have been extensively studied. [0003] Among these methods, because hyperspectral image technology has the advantages of map-spectrum integration, which can simultaneously obtain the geometric morphological...

AI Technical Summary

Problems solved by technology

Although hyperspectral image technology can obtain the geometric morphological and spectral characteristics of seeds, most of the existing research on seed variety identification using hyperspectral image technology uses a single feature of seed hyperspectral image, and there are classification characteristics. Possibility of information loss, leading to underutilization of the advantages of hyperspectral imagery technology

On the other hand, the large number of hyperspectral image bands brings difficulties to its online real-time application in seed variety identification

[0004] Although some band selection methods, such as Uninformative Variable Eliminate (UVE), Successive Projections Algorithm (SPA), Competitive Adaptive Reweighted Sampling Algorithm (CARS), have been applied to the bands of the seed hyperspectral image In selection, but these methods are essentially a feature selection method. When extracting multiple features under each band, it is usually necessary to convert the feature set selected by the above method into a band set, and perform the necessary band intersection and union processing , there is a problem that the accuracy of model recognition is difficult to guarantee

Images