An on-line rule based adaptive computer vision system to control a wood fiberising machine

UK Hardboard manufactures seven different types of board. Each board type exhibits different strength and other end product properties. To obtain these wood fibres with different characteristics are required. Wood fibres are produced in a `refiner´. Currently, the operator visually inspects the fibre quality, and alters the refiner settings accordingly. This was shown to be an inadequate control method. Computer vision techniques have been used to produce repeatable, quantitative measurements of wood fibres. These measurements have then been fed to a rule base which produces controlling actions to the refiner. The system keeps the fibre quality in trim by altering the disc gap. The image analysis results from the planned experiment have shown that the same alterations in fibre properties can be made by adjusting the transport screw speed or the blow valve. An adaptive controller is envisaged