A rule based inspection for printed circuit boards

The authors present a method to recognize and identify various patterns in a scene. The technique utilizes standard design parameters available from reference parameters stored in the database, and it is composed of three main steps. The first consists of properly registering and identifying individual regions in the scene. Each region corresponds to a meaningful object or subobject. Once regions are properly identified, false alarms caused by slight rotation, translation or scaling can be avoided. The second step is to define rules for checking defects. A criterion that governs the appearance of different regions in the image is formed, and algorithms which are capable of verifying the test-parameters and detecting any violations are developed. In the third step, the processing algorithm proceeds by applying the decision criterion to each region in the image and detecting possible defects. Bridging and short-circuit faults for the printed boards are discussed