Today, image processing and machine vision are used in many industrial tasks including process control, monitoring, pattern classification, and object recognition. In an industrial system based on image processing, an imaging device such as a CCD camera is used as the sensing element. The camera provides an image (picture) to an image processor of a scene related to the industrial process (the measurement). The computed results from the image processor are used to determine the necessary information about the process (plant).
A digital camera has an image plate consisting of a matrix of MOSFET elements. The electrical charge that is held by each MOSFET element is proportional to the intensity of light falling on the element. The output circuit of the camera has a charge-amplifier-like device (capacitor-coupled), which is supplied by each MOSFET element. The MOSFET element that is to be connected to the output circuit at a given instant is determined by the control logic, which systematically scans the matrix of MOSFET elements. The capacitor circuit provides a voltage that is proportional to the charge in each MOSFET element.
(a) Draw a schematic diagram for a process monitoring system based on machine vision, which uses a CCD camera. Indicate the necessary signal modification operations at various stages in the monitoring loop, showing whether analog filters, amplifiers, ADC, and DAC are needed and if so, at which locations.
An image may be divided into pixels (or picture elements) for representation and subsequent processing. A pixel has a well-defined coordinate location in the picture frame, relative to some reference coordinate frame. In a CCD camera, the number of pixels per image frame is equal to the number of CCD elements in the image plate. The information carried by a pixel (in addition to its location) is the photointensity (or gray level) at the image location. This number has to be expressed in the digital form (using a certain number of bits) for digital image processing. The need for very large data-handling rates is a serious limitation on a real-time controller that uses machine vision.
(b) Consider a CCD image frame of the size 488 × 380 pixels. The refresh rate of the picture frame is 30 frames/s. If 8 bits are needed to represent the gray level of each pixel, what is the associated data (baud) rate?
(c) Discuss whether you prefer hardware processing or programmable software-based processing in a process monitoring system based on machine vision.