Detection method of fruit appearance based on spectral image analysis

Abstract

The invention discloses a fruit appearance detection method based on spectrum image analysis. The fruit appearance detection method comprises the following steps: pre-determining characteristic wavelengths adopted for carrying out spectrum image analysis detection on a to-be-detected fruit; detecting the outer surface of the fruit by utilizing the characteristic wavelengths through a spectrum image collecting system to respectively acquire spectrum component images corresponding to the characteristic wavelengths, and carrying out principal component analysis on all the spectrum component images to obtain a preliminary test analysis chart; pre-processing the preliminary test analysis chart to obtain a to-be-segmented image, segmenting the pre-processed to-be-segmented image by adopting a self-adaptive threshold segmenting method, and completely segmenting out the image of the slightly damaged area of the fruit in the to-be-segmented image, so as to realize detection recognition on the outer surface of the fruit. According to the method, various slight damages of the outer surface of the fruit can be effectively recognized and the detection recognition accuracy is relatively high.

Description

technical field [0001] The invention relates to an image recognition technology, in particular to a method for detecting fruit appearance based on spectral image analysis. Background technique [0002] Visible light image detection technology can solve the detection problem of significant surface defects of fruits, but the ability to detect minor damages on fruit surfaces such as bruises is relatively limited. The near-infrared spectral analysis technology has good sensitivity and detection accuracy in the detection of internal defects and slight damage, but the near-infrared spectral analysis belongs to point measurement, and there is a strong randomness in the location of the slight damage on the winter jujube, so the detection accuracy Difficult to guarantee. [0003] At present, there are papers or monographs disclosing the detection technology of hyperspectral image technology on the outer surface of fruit, for example, Journal of Agricultural Machinery, 2008, 39(1): 1...

AI Technical Summary

Problems solved by technology

[0004] Although the technical solutions disclosed in the above papers have verified the feasibility of hyperspectral imaging analysis technology in fruit damage detection, there are still the following technical defects: performing Principal Components Analysis (PCA) transformation on all hyperspectral component images, Although the dimension reduction speed is fast, due to too many wavelengths participating in the transformation, it is easy to introduce a lot of noise and reduce the effect of PCA transformation, thus affecting the recognition and detection accuracy.

Images