AUTOMATED DETECTION OF WEEDS AND EVALUATION OF CROP SPROUTS QUALITY BASED ON RGB IMAGES

Abstract

In this paper, we propose a method of automated data processing allowing to detect weeds and assess crop sprouts quality and quantity based on RGB images obtained by unmanned aerial vehicles (UAVs). The process consists of four main stages: 1) vegetation map generation with the use of modified Triangular Greenness Index (TGI); the index is defined as the area of a triangle formed by 3 points on a spectral curve with wavelengths of 480, 550 and 670 nm and estimates leaf chlorophyll content based on RGB images; 2) determination of the position of crop rows and spaces between rows based on the vegetation map; 3) detection of weeds and generation of an appropriate weed map; 4) division of crop rows into non-intersecting fragments and calculating vegetation density in each (the ratio of vegetation area to the total fragment area). By changing the empirically defined parameters of map thresholds of fragment density, one can obtain a map that describes quality of crop sprouts. Unlike existing methods, the proposed scheme does not require presence of infrared data and can be applied to usual RGB images with the use of wide-spread types of UAVs. The method was tested on RGB images of flax and sunflower sprouts collected with SONY ILCE6000 camera in June, 2017 in Altai Territory. The images were taken at the height of 150 m, spatial resolution was 1.5 cm/pixel. The size of each image was 6000x4000 pixels. Test results confirmed high efficiency of the proposed method.

Plant phenotyping relevance

RGB画像から作物の発芽個体群の品質・量を評価する画像解析手法を提案・検証しており、植物状態の抽出が中心的です。雑草検出も含みますが、作物の植生密度による品質評価が明示されています。

Code and data availability

The article describes a TGI-based weed detection and crop sprouts quality method tested on UAV RGB images of flax and sunflower, but contains no data availability statement, no public dataset or image deposit, and no code release. The only URL present (agribotix blog) is a cited third-party reference about vegetation指数

No evidence-backed public reproduction asset is currently recorded.