Browse Prior Art Database

Method for quantifying crop health using overhead imagery

IP.com Disclosure Number: IPCOM000240152D
Publication Date: 2015-Jan-07
Document File: 7 page(s) / 4M

Publishing Venue

The IP.com Prior Art Database

Related People

Ryan Louie: AUTHOR [+3]

Abstract

Measuring agronomic variables such as crop stress and soil fertility is a costly and error-prone process. Overhead imagery can be used to make these measurements less expensively and with greater confidence. Four known analysis tools can be predictably combined to evaluate the image data. These are: 1) the use of a normalized NRGB (near infra-red, red, green, blue) feature space for agronomic purposes, 2) the use of the SLIC (simple linear iterative clustering) algorithm to create superpixels for agronomic purposes, 3) the use of image sets to learn to classify agronomic zones, and 4) the use of multiple levels of clustering for creating agronomic zones.

This text was extracted from a Microsoft Word document.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 29% of the total text.

Method for quantifying crop health using overhead imagery

Ryan Louie, Nicole Rifkin, Brandon Rohrer

Abstract

Measuring agronomic variables such as crop stress and soil fertility is a costly and error-prone process. Overhead imagery can be used to make these measurements less expensively and with greater confidence.  Four known analysis tools can be predictably combined to evaluate the image data.  These are: 1) the use of a normalized NRGB (near infra-red, red, green, blue) feature space for agronomic purposes, 2) the use of the SLIC (simple linear iterative clustering) algorithm to create superpixels for agronomic purposes, 3) the use of image sets to learn to classify agronomic zones, and 4) the use of multiple levels of clustering for creating agronomic zones.

Introduction

The ability to robustly estimate crop health, coupled with other measurements (such as soil nitrogen and yield)is useful in precision farming applications.

Method

Several image processing and machine learning tools can be combined to estimate crop health from images. These are described in the four sections below.

1.       Normalized NRGB feature space

The overhead images are typically composed of four channels, red (R), green (G), blue (B), and near infrared (N). The same region of interest may have very different R, G, B, and N values, depending on lighting conditions, clouds, time of day, and the camera used. Color channels can be easily modified to get normalized red (nR), normalized green (nG), normalized blue (nB), and normalized near infrared (nN). The normalized channels are nearly invariant, allowing the reliable measurement of the relative prevalence of each color, regardless of the lighting, shadows, or time of day. (See Figure 1.)

A normalization scheme similar to that of Kebapci et al. (2010) is used, but extended to include a near infra-red band. The normalized pixel values are calculated as follows:

nR = R / (R + G + B + N)                                                                                                                                            (1)

nG = G / (R + G + B + N)                                                                                                                                           (2)

nB = B / (R + G + B + N)                                                                                                                                            (3)

nN = N / (R + G + B + N)                                                                     ...