National Instruments IMAQTM 用户手册

© National Instruments Corporation

This chapter describes how to perform many common machine vision 
inspection tasks. The most common inspection tasks are detecting the 
presence or absence of parts in an image and measuring the dimensions 
of parts to see if they meet specifications.

Measurements are based on characteristic features of the object represented 
in the image. Image processing algorithms traditionally classify the type 
of information contained in an image as edges, surfaces and textures, or 
patterns. Different types of machine vision algorithms leverage and extract 
one or more types of information.

Edge detectors and derivative techniques—such as rakes, concentric rakes, 
and spokes—use edges represented in the image. They locate, with high 
accuracy, the position of the edge of an object in the image. For example, 
you can a technique called clamping, which uses the edge location to 
measure the width of the part. You can combine multiple edge locations 
to compute intersection points, projections, circles, or ellipse fits.

Pattern matching algorithms use edges and patterns. Pattern matching can 
locate with very high accuracy the position of fiducials or characteristic 
features of the part under inspection. Those locations can then be combined 
to compute lengths, angles, and other object measurements.

The robustness of the measurement relies on the stability of the image 
acquisition conditions. Sensor resolution, lighting, optics, vibration 
control, part fixture, and general environment are key components of the 
imaging setup. All the elements of the image acquisition chain directly 
affect the accuracy of the measurements.

Figure 5-1 illustrates the basic steps involved in performing machine 
vision.