REAL-TIME BLOB-WISE SUGAR BEETS VS WEEDS CLASSIFICATION FOR MONITORING FIELDS USING CONVOLUTIONAL NEURAL NETWORKS

Sky imager cameras provide a full view of the sky with high spatial and temporal resolution and are used to derive cloud cover, cloud type or cloud-base height, if employed in a stereo configuration. The application of a dense fisheye stereo method provides dense, consistent and quite complete 3-D cloud boundaries and can be fully automated. We present validation of our approach and cloud examples with high geometric complexity. Applications are radiative closure studies and cloud dynamics.