Categorization of objects in image processing

The image processing representing road scenes introduces a very important volume of information and requires a substantial elaboration at all levels: processing, Segmentation, Recognition and Interpretation. It influences unmistakably the quality of the tasks of processing and the run time. But, the criteria quality and time are often the only ones retained by the applications, in particular in real time. Hard operations in processing such as the detection and the recognition of objects require an adequate internal representation of the systems of vision embarked on vehicles: Plans of structuring were proposed in this representation; we retain here those who consider the operation of recognition in two steps: Categorization and Identification. The categorization prepares a number restricted by categories of objects considered relevant completing the information for the task of the identification. We propose on this base a procedure of categorization of objects using an evaluation capable of adjusting the level of division to facilitate the recognition.