Rice canopy recognition method based on digital camera image

Abstract

The invention relates to the field of crop canopy image recognition, particularly to a rice canopy recognition method based on a digital camera image. The method comprises: (1), 1) obtaining a by using a digital camera; (2), obtaining color values of R, G, and B channels in the color rice canopy image; and (3), determining whether the image is a rice canopy according to the R, G, and B color values in each pixel of the color rice canopy image. Therefore, the identification rate of the method is high; a working step of selecting an image sample in advance according to the existing common method can be removed; the calculation process is simple; and the work efficiency can be improved obviously.

Description

[0001] 【Technical field】 [0002] The invention relates to the field of crop canopy image recognition, in particular to a rice canopy recognition method based on digital camera images. [0003] 【Background technique】 [0004] Crop canopy image recognition usually refers to identifying the canopy part of green plants in the acquired satellite remote sensing images, camera images, and scanner images, so as to eliminate the non-canopy parts (such as soil, water, dead leaves, shadows, etc.) Wait). Crop canopy image recognition is mostly used to estimate plant coverage and leaf area index, and then evaluate crop growth, estimate yield, etc.; or for rapid diagnosis of crop nutrition. In recent years, with the promotion of digital cameras, it is very convenient to use digital cameras to obtain canopy images. Therefore, research on crop canopy recognition using digital cameras has considerable application prospects. [0005] There are three main methods of canopy identification curre...

AI Technical Summary

Problems solved by technology

[0005] There are three main methods of canopy identification currently used. One is to use the "color selection" program of Photoshop software to manually select the canopy area in the image, which has the disadvantages of time-consuming, labor-intensive, and uncertain human errors.

The second is supervised learning classification methods, such as support vector machine, neural network model, maximum likelihood method, etc. These methods need to manually select pure image samples for supervised learning classification, and then learn according to the samples to determine the classification basis; The disadvantage is that the calculation time is long, and the image sample needs to be selected in advance, and the operation is not convenient enough

The third is to use the color values ​​of the red (R), green (G), and blue (B) channels in the image, according to the spectral reflection law of the plant canopy, set the corresponding calculation formula to distinguish; this kind of method is more commonly used The index is: 2g-r-b, where r=R / (R+G+B), g=G / (R+G+B), b=B / (R+G+B); the classification is based on the image The 2g-r-b difference between soil and green plants is obvious, and the image binarization can be used to distinguish soil and green plants; the disadvantage of this method is that the binarization threshold needs to be set according to different images, and images can be compared under different shooting conditions. The threshold setting has a great influence on the classification results. It is easy to misclassify the leaves covered by shadows at the bottom as canopy leaves, and at the same time classify the canopy leaves under strong light as non-canopy leaves. The classification accuracy is usually not as good as The first two methods

The existing canopy identification methods have human interference factors and complex calculations

Images