Warning: A non-numeric value encountered in /homepages/29/d761004148/htdocs/clickandbuilds/KermitImageToolkit/wp-content/themes/Divi/functions.php on line 5560

Dataset Library

by | Feb 27, 2017 |

This page lists the datasets we have made available in the past years.

Title: Lopez-Maestresalas 01 (LMO1)

Author: A. Lopez-Maestresalas.

  • This dataset was first introduced in “Baddeley’s Delta metric for local contrast computation in hyperspectral imagery” (more info).

Contents: 40 hyperspectral images (320 columns, 80-160 rows, 256 spectra), toghether with its ground truth for object-from-background segmentation. The former are distributed as *.mat files, while the latter come as *-png binary files.

Specification: The dataset is composed of 40 hyperspectral images of 320 columns and a variable number of rows (from 60 to 180 rows). At each pixel, the spectra information is measured at 256 different wavelengths with 16-bit precision. For our experiments, such precision has been limited to 100 discrete reflectance levels. Each of such images features one or more items to be segmented from the background.

The vegetal material used in this study for experimental validation consisted in a set of fruit and vegetables manually collected from the Agricultural Practical Work and Research Facility of our own University during October 2016. Each piece was individually scanned through a hyperspectral imaging system composed of a hyperspectral camera, an illumination unit, a transportation plate, and a standard computer. The model of hyperspectral camera is Xenics Xeva 1.7-320, featuring an InGaAs sensor and a resolution of 320×256 pixels;     the camera comes coupled to an imaging spectrograph, (Specim Imspector N17E). Both devices have sensitivity ranging from 900 to 1700 nm. Illumination of the samples was achieved with four 46W AC halogen lamps placed above the transportation unit. The system acquires images of one pixel height and 320 pixels width, where the {x} axis is a spatial coordinate and the {y} is the spectral one, that represents the reflectance value of each pixel at the different wavelengths along the 900-1700 nm spectral range, every 3 nm.

Cite as: A. Lopez-Maestresalas, Dataset LMo1, December 2016.

Link for download: Distributed under demand by A. Lopez-Maestresalas (ainara.lopez-at-unavarra.es).