The speckle reducing on the multidimensional television signals by two-camera profile sensor

The speckle reduction on the multidimensional television signals is an important task in measurement machine vision systems. The greatest error due to speckle is detected in triangulation optical sensors if measuring objects have a periodic structure. When measuring such objects, a random interference pattern determines speckle noise. Speckle noise can be reduced by a two-camera profile sensor. Cameras capture the multidimensional image of an object in different angles and the interference pattern on the images will also be different. This allows removing it from the image. The first feature of processing is signal superposition in conditions of distortion. This problem is solved by preprocessing. The second problem is speed processing of superposition. It is solved using pyramid transformation and optical flow estimation. The developed speckle noise reducing technique was tested on multidimensional television signals with images of drill-pipe threads. In the comparison of single-camera profile sensors the error of the shape estimation of the object decreased from 0.20 mm to 0.05 mm.