1 August 1990 Real-time colour classification for preprocessing photogrammetry images
Author Affiliations +
Proceedings Volume 1395, Close-Range Photogrammetry Meets Machine Vision; 139511 (1990) https://doi.org/10.1117/12.2294280
Event: Close-Range Photogrammetry Meets Machine Vision, 1990, Zurich, Switzerland
Abstract
The automatic processing of stereo images for 3D information extraction can be simplified by colour classification of the stereo image pairs prior to the correlation matching procedures. The number of corresponding pixels to be matched is drastically reduced, if only those pixels on the epipolar line are correlated which belong to the same class of colours. Colour classes of natural scenes like aerial photographies are usually composed of complicatedly shaped clusters which makes the manual setting of RGB classifier parameters an almost impossible task. We present a real-time IHS-colour classification system which uses neural network principles based on self-organizing look-up-tables for learning typical colour classes . A learning rule for the supervised training of this LUT classifier is presented . The proposed LUT classifier shows all the positive features of a 3-layer perceptron Neural Network, but performs 70.000 times faster then a simulated perceptron network and uses low-cost, commercially avai- lable components. The height measurement of plantlets for the automation in greenhouses is presented as a typical application of these principles.
© (1990) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Robert Massen, Robert Massen, Gerhard Volk, Gerhard Volk, } "Real-time colour classification for preprocessing photogrammetry images", Proc. SPIE 1395, Close-Range Photogrammetry Meets Machine Vision, 139511 (1 August 1990); doi: 10.1117/12.2294280; https://doi.org/10.1117/12.2294280
PROCEEDINGS
8 PAGES


SHARE
Back to Top