A Test for the Systematic Detection of Bowtie Effect and Stripping on the Edges of MODIS Images

MODIS images are distorted at the edges due to the nature of the scanning array (whiskbroom scanner). This distortion, known as bowtie effect, is produced because the geometry of the acquisition and the sensor array design causes overlapping of pixels on the edges of the image. Those pixels are larger than the ones at the centre, causing the image have a bowtie shape. To tackle this issue, in this work we have produced a test to detect the bowtie effect and filter the MODIS images that are affected by it. We have applied the test to a subset of tiles of images from the CUMULO dataset, producing a filtered subset of images. We have then built a dataset with images filtered from a ran of the test and uploaded it into a public database, in Kaggle. The test developed can be used to detect bowtie and periodic stripping defects in any image, useful to filter datasets for machine learning training.