Method for identifying barnyard grass of rice field by utilizing a hyperspectral imaging technology

Abstract

The invention discloses a method for identifying barnyard grass of a rice field by utilizing a hyperspectral imaging technology, and belongs to the technical field of identification and early warningof noxious weeds in rice fields. The method comprises the steps of planting germplasm resources of rice and barnyard grass; selecting the rice and the barnyard grass, and acquiring spectral data through a hyperspectral imaging system; taking the spectral data as a model input variable, performing image calibration and image processing through computer software to obtain the average reflectivity ofthe whole leaf, and after signal denoising, selecting a special spectrum to build a discrimination model; and carrying hyperspectral imaging equipment by an unmanned aerial vehicle to collect spectral data of a large-area rice zone, and finally calculating the occurrence rate of the barnyard grass to evaluate the harm degree. The barnyard grass of the rice field is identified by adopting the hyperspectral imaging technology; by utilizing a difference characteristic of the rice and the barnyard grass, the barnyard grass and the rice are distinguished; six main wave spectral characteristics arescreened, so that the calculation is reduced and the accuracy is up to 98.1%; and an unmanned aerial vehicle carrying technology can be used for carrying out high-altitude operation, so that a spectral data result can be rapidly and non-destructively acquired in a large-area manner.

Description

technical field [0001] The invention belongs to the technical field of identification and early warning of malignant weeds in paddy fields, and in particular relates to a method for identifying barnyard grass in paddy fields using hyperspectral imaging technology. Background technique [0002] Barnyardgrass is a worldwide malignant weed, and it is one of the main weeds with the most occurrence, the most serious damage and the widest distribution in rice fields. Barnyardgrass can not only compete for water, fertilizer, light and growth space, reduce the yield and quality of rice, but also be the host of planthoppers, rice stink bugs and other diseases and insect pests, seriously endangering the production safety of rice. In recent years, due to the widespread application of rice direct-seeding technology and precision-hole direct-seeding technology, barnyardgrass damage has become more rampant, leading to the phenomenon of rice yield reduction or even failure. Early identifi...

AI Technical Summary

Problems solved by technology

[0003] However, in actual rice production, the identification of barnyardgrass in paddy fields is still in the backward stage of morphological identification. Judging the incidence of barnyardgrass in paddy fields based on the difference between auricles and ligules is still a commonly used method for large-scale rice cultivation in large farms.

Barnyardgrass and rice share a common evolutionary origin, and have similar growth phenotypes at the 3-4 leaf stage (the key grass-controlling stage). Subjective evaluation of the number and damage of barnyardgrass often has a large deviation in the actual production of rice. Untimely prevention and control or excessive application of herbicides

The second method is the RGB color difference method to identify the characteristics of weeds. This method has the problem of low recognition rate. At the same time, the accuracy of this technology is easily affected by the environment and cannot be popularized. The third method is hyperspectral imaging technology. The technology can obtain spatial information and spectral information at the same time, but due to the large amount of data, serious data redundancy, and large computer memory occupation, it was mainly used in laboratory offline analysis in the past.

In 2011, American scientists used visible / near-infrared multi-spectral to identify associated weeds in corn, with an accuracy rate of up to 96.7%, but there is still a problem of serious data redundancy that cannot be applied

In 2015, the 11-dimensional (1000-2500nm) band was used to identify weeds in the cabbage field, and the overall accurate identification rate reached 96.8%. Although the amount of calculation was greatly reduced, the research was mainly aimed at inter-row identification and could not be applied to the identification of barnyardgrass in rice fields.

[0005] In 2014, the artificial neural network classification method was adopted and set to 7 main spectral features, which can distinguish wild oats and red root hogweed in pea fields, spring wheat fields, and rapeseed fields, with an accuracy rate of 88-94%, but the advantages of this method There are only 7 bands, and the amount of calculation is reduced, but the disadvantage is that the recognition rate is low. It is installed and fixed on a flatbed truck with telescopic arms. It is suitable for dry land operations and is not suitable for large-scale rice fields.

Images