Non-contact surface roughness measurement based on laser technology and neural network

A non-contact surface roughness measurement method based on laser speckle, image processing and neural network technique is introduced. As the laser speckle patterns contain a lot of information about illuminated surface, four feature vectors correlative to surface roughness, which include contrast, dark region ratio, gray distribution and binary feature, are extracted and taken as inputs of the neural network to realize the surface roughness measurement. Neural network has characteristics, such as automatically organizing, automatically studying and memory capability etc; therefore, after training the network by a number of examples, the measurement can be implemented. 4 flat-grinding specimens with different roughness values are measured in the experiments. The results indicate that the measurement method has the advantages of not-contact, fast, precise, simple and easy to implement.