Matching is one of the most used algorithms in image processing. The basic point is to teach the original picture (usually a specific part) and look for this original part in other pictures taken.

Matching is widely used in almost every field of machine vision such as measurement, error detection, completeness check, print monitoring, positioning, direction finding, 3D, and Robotics.

Matching generally gives us a percentage.(95% matched, 76% matched). In addition, it specifies the coordinates of the matching part and the angle of rotation. It is often preferred by image processing developers because it generally requires less coding.

Matching consists of;

Transformation: Perception of angular and dimensional change. The objects to be recognized may be closer or further to the camera, or it may come in much different positions than the center or the original picture. In such cases, the original part needs to be transformed into the taken image. This process is called transformation.

Alignment: If there are any reference points in the taught example, these points must be placed in their place.

Rectification: The process of bringing the picture found at different angles and positions to the same angle and position as the taught original one

In the example picture, only a specific part is taught. Along with being able to teach the whole picture, the part which you want to capture in the image is taught. In this example, ViViDent is taught if we want to capture ViviDent among many different brands. Among ViviDents, if the "NANE aroma" (means Minty) is to be searched, the bottom "NANE aroma" is taught.

All products, center coordinates and angles of rotation can be captured with matching at different angles, even on top of each other.
In some applications, matching is used for purposes such as catching the faulty product or detecting a different product. In some applications, it is used only as an intermediate step. Once you have found the product, you can take measurements on the product, take it with the robot, count them etc.

 

3D MATCHING

It is also possible to match images via 3D cameras or 3D CAD drawings. By means of this technique, also known as Surface Inspection, it is possible to catch defects on the surface in 3D.

Solutions we use matching as AKU

• Counting Applications: Counting products on the conveyor or on the surface by camera
• Centering, Positioning: Determining the center of the products coming from different angles and the amount of rotation and taken or fed by robot or automation system.
• Fault detection: According to the fully taught product, the detection of deficiencies, printing disorders, formal problems
• Measurement: Making desired measurements of the product from different angles and positions
• Recognition: To distinguish each product among different types

 

When used in conjunction with deep learning, matching is challenging in many different areas of use. Identification of vegetables or fruit for automatic pricing in weighing process, identification of books from cover images without barcode reading etc.