Automatic classification system of soil cover based on color information

This work considers the problem of using a digital color image analysis to determine the percentage of soil cover. A triplet is associated to each ima ge element representing a color in a three-dimensional color space of the t hree basic colors: red, green and blue. After the image acquisition, sample s were taken from the images to define the standard colors of the soil cove r and soil. By the Hotteling Transform, was obtained three new components n ot correlated, and histograms based on its occurrences in samples were anal yzed. Once the patterns of soil cover and soil were established the classif ication algorithm labels the image pixels according to its corresponding pa ttern. After that, the soil cover is visualized and quantified. A range of crops have been studied, verifying that they achieved 60% coverage in about 65 days. Several runs were performed to check the precision of the tool, a nd results were compared with the corresponding conventional techniques. Th e software to be presented allows to estimate soil cover in a fast, more pr ecise and cheap way compared to the conventional ones.