Line Scan Cameras
A line scan camera is very similar to a flat bed scanner, containing a line of single light sensitive elements. Some line scan cameras can contain several lines of elements but the light values would be integrated to form a single line of data so the processing mechanism is the same. As it is only a thin line of sensing elements, the resolution is often very high, commonly 1024, 2048 or 4096 pixels per line with higher resolutions available. Applications suited to line scan cameras are often those of moving objects but are also used for large objects, supervising continuous production processes or for inspection of circular objects which can be rotated in front of the camera.
Line scan cameras are normally categorised by their sensor length and pixel clock rate. The sensor length simply defines the number of sensing elements i.e., pixels it contains. This factor ultimately dictates both the physical length of the sensor and the camera lens size. Getting the lens size right is an important factor in line scan cameras. If the lens is too small, intensity variation may result towards the outside of the lens.
This is known as vignetting and its effects are shown in the following diagram:
The pixel clock rate is the rate at which the pixel values are read from the camera which is much higher in line scan cameras than area scan cameras. This rate can go well beyond 100MHz and as such line scan cameras are typically digital as the analogue bandwidth can be too limiting.
Processing a line scan image is very different to processing an image from an area scan camera as the image is not delivered in a rectangular form.
There are three main processing methods as follows:-
This method processes only the current line of data at any time and for this reason it is suited to tasks such as detecting the presence of a part, irregular surface abnormalities such as holes or if the inspection is merely a formation of light or dark lines. This type of processing often is done in camera and results in low-cost and small overall size. However these systems can be laborious to setup and the limited in camera processing can hinder the overall application complexity.
Unlike the fore mentioned line-wise processing, this method takes a succession of lines from the camera and combines them to form a rectangular image. The control of this method is usually signalled by a trigger to begin and collated over a predetermined number of lines. This means the image should always start and end at the same place. The line frequency or rate of image acquisition must be accurately generated. This method is commonly used to achieve a total surface image of a cylindrical object which is rotated in front of the camera and lines are collated to form a net image as if the surface were rolled flat.
A combination of the two previous methods results in this most expensive and complex process. A rectangular image is formed as in piece-wise processing but is updated with every newly scanned line as in line-wise processing. The processor which supports this method would need to be able to append the recent line from the camera and discard the oldest using cyclic shift registers. To ensure a rectangular image is ready for use at any one time, the processing rate must match the scan rate where the image is evaluated before each acquisition of a new line. The expense of this method is generally high as the memory bandwidth and processing requirements often result in the application of parallel computers.
When a building up an image of a moving object, it is important that the speed of movement matches that of the sampling rate. Normally this is achieved by using an encoder to advance the moving part (camera or object) by a uniform amount.
Complexity in designing and using line scan camera applications arises from the lack of a common standard in the field.