Research On Intelligent Detection Technology For Surface Defects Of High Reflective Parts

Machine vision has been widely used in non-destructive testing because of its high efficiency and non-contact.But if the light is too strong,the small scratches and pits are easily covered by the reflection of the surface,which makes it difficult to extract the defect features.When the illumination condition is enough to make the defect features present in the image,other non-defect features of the surface,such as dust,water stains,and fine lines on the surface caused by the vibration of milling tools are inevitably regarded as defective features,resulting in the misjudgment of qualified parts.In view of the above problems,the following research work has been completed.1.Aiming at the high reflective characteristics of the tested parts and the information of the shape and size of the defects that may appear on the tested surface,the problems that may occur in the process of testing and the corresponding solutions are analyzed in depth.According to the emphasis of detection task,the requirement of hardware equipment and the difficulty of software algorithm,the detection scheme is designed and selected.A machine vision defect detection method based on image processing and depth learning is proposed.2.The high reflective surface defect location method is studied,and the defect location model is designed and tested by using image processing algorithm,so as to achieve 100% defect area detection rate and obtain a location model with low false detection rate.3.The neural network algorithm model is designed to discriminate the features of the local image of the candidate region determined by the location model.The model can effectively filter out the false detection region produced by the location model and has the ability of generalization.4.Design and build an image acquisition and detection system.The circular diffuse reflection illumination method is adopted to ensure that the defect features are clearly presented in the image,and the interference caused by the reflection will be minimized;the detection program is compiled according to the designed detection algorithm flow and principle,which has the functions of automatic snapping,automatic defect area acquisition and storage;the validity of the method and various items are verified.Performance test results show that the method can meet the detection requirements of the detection accuracy,and has the ability to generalize.