A Method of Cubic Object Feature Extraction

Access this Article

Search this Article

Abstract

How to reduce and simplify the calculation for image recognition is a very attractive and important issue in order to realize the real time control of a robot based on the image recognition results. This paper describes a method of extracting 2 - dimensional geometrical features of cubic objects based on the normal vector distributions from the visual information obtained with the laser range finder to reduce the calculation of the image recognition. In this research a laser beam is scanned in the horizontal plane to which the cubic objects stand vertically and the laser spot is detected with a TV camera every sampling time. These spots make an intermittent locus which includes some special lines corresponding to the cubic objects. To extract the features of the cubic objects, we utilize the normal vectors formed on the locus. If some normal vectors distribute in the same direction and the origin of the normal vectors are very close to their neighbor's, these normal vectors can be classified into the same class, -the straight line class. Because the normal vectors on the neighbor surfaces of the cubic objects are vertical to each other, we use this property to determine the pair of straight lines which belong to the cubic objects. Making the histogram based on the normal vectors with the same direction, we obtain the peaks which are supported by the points on the cubic object surfaces. Then, the points can be extracted from the set of points on the whole locus inversely according to the relations with the peaks and the features of the cubic object can be extracted by applying method of least square to these extracted points. The experiments proved the availability of the proposed processing algorithm.

Journal

  • Memoirs of the Faculty of Engineering, Okayama University

    Memoirs of the Faculty of Engineering, Okayama University 25(1), 1-13, 1990-12-14

    Faculty of Engineering, Okayama University

Codes

  • NII Article ID (NAID)
    120002308010
  • NII NACSIS-CAT ID (NCID)
    AA10699856
  • Text Lang
    ENG
  • Article Type
    departmental bulletin paper
  • ISSN
    0475-0071
  • Data Source
    IR 
Page Top