Surface Roughness Detection Based On Highlight Feature And Its Application

When light hits the object,the object surface produces bright spots of different sizes and shapes densely distributed in a certain area of the object surface.The spot area is called a highlight region.The highlight is usually treated as a noise screen due to local color distortion that the highlight is lost.When the contactless class detects the surface roughness,the surface of the object thus affects the detection results.But the highlight is different from overexposure,and can also reflect the surface roughness through the form of brightness shock.This article defines highlights as a local feature that can reflect the uneven information on the surface of an object.Compared with three classic lighting models,the highlight illumination model for the local characteristic is established as Highlight model.The Highlight model is extracted using the combination of least squares curve fitting and wavelet transform.By analyzing the relationship between highlight features and surface roughness,a new surface roughness detection method is proposed.The highlight features are applied to surface roughness detection.Experiments on ceramic tiles demonstrate the effectiveness of surface roughness detection based on highlight features.And based on the highlight characteristics of different materials,the highlight features also have a certain research and application value for object recognition.A highlight image acquisition platform has been built to adjust the relevant parameters of the lighting environment and the image acquisition device to ensure that the highlight images can be accurately and clearly collected,and the images are not overexposed.Highlight features are extracted from the collected highlight images to detect the object surface roughness and object recognition based on the highlight features.Specific research contents are as follows:1.Analyze the principle and application scenarios of three classic lighting models: Lambertian model,Phong model and Rendering Equation model.A Highlight model based on the dark room and the light source is the only direct source is presented,through which it highlights the local feature of highlight.2.Target the proposed Highlight model.The combination of least squares curve fitting and wavelet transform is characterized as a bright and dark oscillation trend in the frequency domain.Several materials are selected to test the highlight detection algorithm.3.The highlight features are further described to obtain the highlight eigenvalue Z.The surface roughness of different tiles was detected by highlight eigenvalue Z;based on highlight eigenvalue Z,the convolutional neural network(CNN,Convolutional Neural Networks)was built and based on highlight feature identification,the identification rate reached 82%.