Images from the camera are digitized and stored in the computer where it is automatically analyzed to extract the required information. The camera can either be a standard video camera, or in the cases of products on a conveyor, a linear array sensor which generates a high resolution image on a line-by-line basis. It is usually not necessary to store an ideal image of the item being inspected. Rather, certain features of a correct item are combined to form a model of an ideal product. The same features are extracted from each inspected item and compared with the model. In this way factors such as positional alignment, changes in illumination levels and tolerances in specifications are accommodated. Often a conventional computer is not fast enough for real applications. In such cases special hardware must be used for parts of the problem.  This can take the form of custom designed preprocessing hardware or commercially available high-speed vision architectures.



  • Spatial color nonuniformity correction

  • Positions of objects

  • Nonparametric color image segmentation model

  • Feature values describing

  • Position of each object

  • Size of each object

  • Shape of each object

  • Orientation of objects