Area array and line scan cameras are each suited for unique machine vision applications. Area array cameras, for all intents and purposes, are “conventional” cameras that use sensors with two-dimensional pixel arrays. The square or rectangular shaped sensor captures an image in a single pass with the resulting image having a width and height corresponding to the number of pixels on the sensor, for example, 640x480. Because of this, area array cameras are ideal for machine vision tasks where objects are small or have approximately the same size in both dimensions. However, the size of PCBs, LCD panels, and wafers has increased beyond the speed, accuracy, and resolution capabilities of many area array cameras. Line scan cameras offer a better solution.
A line scan camera uses a single row of light-sensitive pixels that image across the object, line-by-line, accompanied by high intensity lighting. Resolution is specified in the horizontal axis since the achievable resolution in the vertical direction will depend on the clock rate of the camera and the speed of the web. A completed image is built by stitching together the lines, much like a fax machine. Because only a one-dimensional correction needs to be applied, line scan cameras are much easier to correct for lens shade, photo response non-uniformity (PRNU), or dark signal non-uniformity (DSNU), than an area array camera.
Line scan pixels accumulate photoelectric charges relative to the light from the object imaged onto that pixel. Next, a readout register amplifies, adjusts, and digitizes the charges, all while the next row of pixels is being exposed. The maximum rate at which exposure and readout can occur is the “line rate,” calculated in kilohertz (kHz) — the number of lines exposed in one second. In production, the faster an object is moving, the higher the required line rate. To avoid under- or over-sampling an object, a programmable encoder, often connected to a conveyor or web, measures speed and precisely synchronizes the camera in pulses. A predetermined number of lines of the image are then stitched together to form a frame that is analyzed with software. Any defects are recorded on roll maps.
Line scan cameras excel at producing a flat image of cylindrical objects, at imaging very large objects with high resolution and at producing images of objects in continuous movement past a fixed point, such as parts on an assembly line or web applications. Line-scan applications include paper, rolls of metal, fiber, railway inspection, solar cells, textiles, pharmaceuticals, semiconductors, and postal sorting. Another advantage is that the cameras can fit into tight spaces, for example when they must see through rollers on a conveyor to acquire images of the bottom of a part.
In certain applications demanding both high scan rates and high contrast however, the sensitivity of line scan cameras using single x1 linear sensors can fall short. Increased sensitivity requires multi-line scan cameras. Dual-line scan designs feature two parallel arrays of pixels, capturing twice the number of photons and doubling sensitivity. To improve sensitivity further, time-delay integration (TDI) is frequently incorporated into line scan cameras. TDI-based cameras have several vertical integration stages, resulting in the capture of multiple exposures of the same object. Integrating the output from these stages increases sensitivity.
Single-line monochrome line scan cameras have linear sensors consisting of multiple pixels in a x1 configuration. To obtain a color image from a single-line scan imager, a linear R-G-B-R-G-B filter can be applied to the sensor with the pixels merged to create a color image. Unfortunately, this approach produces an interpolated image with lower resolution.
A “trilinear” approach calls for each of three arrays to capture one primary color simultaneously but at somewhat different locations on a moving object. The channels are then combined to form a full color image. Spatial correction compensates for the separation — the first and second arrays are buffered to match the third. The downside of using only three channels is relatively low spectral resolution. Manufacturers have improved the performance with image-based color measuring approaches that enable color to be measured on the whole surface of the object, not just on one spot, as with traditional spectrophotometers.
For truly accurate color inspection, line scan cameras with more than three color channels are required. Modern multispectral line scan cameras feature 6 – 12 spectral channels in the 360 – 960 nm range. Multi-channel imaging provides accurate spectral and color output on varying substrates such as paper, plastics, films, and foils.
Color imaging may no longer be enough, however, for inspection where specific wavelengths are required that are either outside the visible spectrum or in between the RGB color bands. Multispectral cameras can be used from near IR up to 960 nm in that case.
3D Line Scan Inspection
Over the past decade, camera manufacturers have introduced several 3D methodologies, ranging from time-of-flight (TOF) analysis and projected pattern correlation, to laser line/triangulation measurements and stereoscopic technologies. Of these, stereo has gained stronger traction, particularly in the semiconductor industry. Components, such as solder balls or pins, which are used to connect wafers and dies, have to be inspected with 3D methods to precisely measure the critical height of the conducting elements. The typical dimensions of such components are about 50 μm, requiring an optical resolution for the inspection systems in the range of at least 5 μm.
The basis of the stereo technology is similar to human vision. Two sensors — in this case, linear sensors — in a stereo configuration are combined into one camera, resulting in two images being acquired of the same object from slightly different perspectives. This serves as the basis for triangulation, which involves an object point projected in both stereo images, and two image points corresponding to the positions of the right and left camera.
Manufacturers are now combining the best of both worlds — line scan with stereo. These cameras have linear sensors up to 8000 pixels in RGB to provide both high resolution and large field of view. Because of the improved accuracy, they open up new 3D applications that are not possible with other approaches in order to detect the most minute of defects. Another advantage is speed — linear sensors have up to 50 kHz line rates even at extremely high resolutions. Finally, this approach results in fewer occlusions — the stereo line scan cameras are oriented perpendicular to the object surface so there are no occlusions in the transport direction.
Compared to manual inspection, machine vision systems employing area-scan cameras offer improved accuracy and far higher consistency. For all their advantages, however, there are limitations to area-scan cameras in more challenging machine vision tasks. Line scan cameras are available today with numerous sensors, speeds, and interfaces so developers can choose the one that best fits their applications.