Fluff separation by image analysis

In demolition processes of the motorcar after dismounting, shredding, metallic materials sorting and after an appropriate venting and air classification, a special fraction is obtained, conventionally named fluff. Fluff is a light fraction composed by organic material, rubbers, tissues, normally used for the stuffing, the insulation, the gaskets and the tapestry of a car. Although such a fraction is not able to be discriminated by specific weight based methods, the separation in its main constituents is required to get recycled products of high commercial value. Particularly it is necessary to get an organic fraction, composed exclusively by plastics and rubber and deprived as far as possible of contamination due to other materials in the fluff mixture. Therefore an optical selection technique based on image processing, aided by proper software of recognition, has been adopted. The system of image survey is based on digital black and white (B&W) image acquisition of the objects that flow and are properly lighted. A suitable number of parameters are obtained for each object. From the gained data, a feature vector of representative objects has been extracted in order to achieve objects selection in 2 classes according to the market specification for recycled materials. The classification of the objects in different classes of destination has been carried out using a classification algorithm adopting various experimental modalities. Satisfactory results in terms of correct recognition (>80%) have been achieved.