The environmental impact from herbicide utilization has been well documented in recent years. The reduction in weed control with out a viable alternative will likely result in decreased per acre production and thus higher unit production cost. The potential for selective herbicide application to reduce herbicide usage and yet maintain adequate weed control has generated significant interest in different forms of remote sensing of agricultural crops. This research evaluated the color co-occurrence texture analysis technique to determine its potential for utilization in crop groundcover identification. A program termed GCVIS (Ground Cover VISion) was developed to control an ATT TARGA 24 frame grabber; and generate HSI color features from the RGB format pixel data, HSI CCM matrices and the co-occurrence texture feature data. A set of 40 test images from six different species of ground cover: crab grass, foxtail, lambsquarter, morning glory, velvetleaf and soil; were digitized using GCVIS. SAS procedures were used to determine the discriminant powers of the texture features and then conduct classification analysis to determine the classification accuracy. The isolated species discriminant analysis proved that color co-occurrence methods have very high classification accuracy for the set of eleven texture features selected by PROC STEPDISC, a 93.33 % classification accuracy was realized for the four ground cover conditions.
Weed Recognition Using Machine Vision and Color Texture Analysis
Sae Technical Papers
International Off-Highway & Powerplant Congress & Exposition ; 1996
1996-08-01
Conference paper
English
The Illumination-Invariant Recognition of Texture in Color Texture
British Library Conference Proceedings | 1995
|Machine Vision-Based High-resolution Weed Mapping and Patch-Sprayer Performance Simulation
British Library Conference Proceedings | 1999
|Address Block Location Using Color and Texture Analysis
British Library Online Contents | 1994
|Color Models for Outdoor Machine Vision
British Library Online Contents | 2002
|Color machine vision advances computerized inspection
British Library Online Contents | 1995
|