Computer Vision News - April 2016

Project Automatic Flowers Classification Every month, Computer Vision News reviews a successful project. Our main purpose is to show how varied image processing applications can be and how the different techniques allow to solve technical challenges and physical difficulties. This month we review some aspects of an Automatic Flowers Classification software developed by RSIP Vision engineers to give flower growers an indispensable tool to compete in the marketplace. Do you have a project in computer vision and image processing? Contact our consultants . COMPUTER VISION NEWS Background 20 million flowers are traded every day at the monumental Aalsmeer Flower Auction near Amsterdam in Holland . Its floor space of nearly one million square meter makes it the fourth largest building in the world. Flower growers from over the world sell their product at the auction, which Dutch people use to call bloemenveiling . In order to get the best price at the auction, sellers need to offer their merchandise classified according to the most precise and objective measurement methods , such as can be developed with computer vision and image processing. In particular, they need to certify the stem width, a reliable indicator of the life expectancy of cut flowers. Solution The algorithms of RSIP Vision operate on pictures taken on a conveyor belt in the factory. One of the challenges which automatic flower classification must overcome is segmenting the flower from the background. In principle, it would be possible to use conveyor belts or container bins with colours contrasting with that of the flower itself, so that the object is easily recognized in the picture. Unfortunately, background cannot be trusted as a reliable input since the conveyor belt and the bin will become dirty and loose the original contrast. As in many precision agriculture applications, rescue comes from infrared technology . Thanks to chlorophyll, edges of plants, leaves and stems are very clearly identified by infrared cameras, enabling thus a precise segmentation of the flower. This is followed by a much higher-resolution picture, taken to identify the colour. 12