Surface Defect Detection Of Automotive Turbine Shell Parts Based On Machine Vision

In order to improve the fuel efficiency of automobiles,save energy and reduce carbon emissions,automobile manufacturers in various countries have adopted engine turbocharging technology.Turbine shell is the key component of turbocharger,and its surface quality directly affects the assembly and service performance of the parts.At present,manual visual inspection is the main method for surface quality inspection of turbine shell parts.Due to the long-term eye fatigue and subjective inconsistency of the criteria,the detection accuracy is poor,the efficiency is low,and the labor cost is high.In recent years,machine vision technology has developed rapidly.With its advantages of high accuracy,efficiency and real-time,it has been widely used in product detection and industrial automation.The machined surface of turbine shell parts can be divided into sealed area and non-sealed area.The surface quality requirements of different areas are different.And because the coolant in the process of processing leaves spots on the surface of the parts,it is difficult to detect the surface defects of the turbine shell parts.In this paper,machine vision technology is applied to defect detection of machined surface of turbine shell parts.Based on the analysis of the characteristics of surface defects of parts and the requirement of inspection technology,the overall scheme of inspection system is formulated.The hardware unit including image acquisition,parts clamping and grabbing,motion detection and so on is constructed.Software modules such as visual inspection and motion control are developed,and defect detection such as scratches and pits on the surface of parts is realized.Visual inspection module is the core of the whole system.In this paper,the principle of illumination compensation is studied,and the signal-to-noise ratio(SNR)of scratches and pits in images is enhanced by the combination of high and low angle lighting.In the scratch detection algorithm,this paper obtains the background image which only contains the speckle feature in the region of interest by morphological median filtering,and then gets the scratch segmentation image by background difference.Next,according to the morphological characteristics of scratches,a region growing algorithm based on directional gradient is used to connect the disconnected scratches.Finally,the weight of each morphological single feature is obtained by confidence analysis,and scratches are identified by multi-feature fusion method.For pit defect detection,the precise boundary extraction algorithm based on ellipse fitting is firstly used to get the region of interest,and then the morphological features are extracted by image segmentation in the pit area.Some of the features are used as input vectors of support vector machine,and machine learning method is used to identify pit defects.According to the test results of Nissan GP12 and Great Wall 1.5 turbine shell parts,the surface defect detection system based on machine vision developed in this paper can detect scratches and pits with a single scale of more than 0.3 mm.The missed detection rate is less than 2%,and the single plane detection time is less than 2 seconds,which can meet the requirements of on-line inspection in industrial field.