Research And Implementation Of Automatic Optical Inspection Algorithm For Apparent Defects Of Magnetic Materials

Due to its special physical properties magnetic materials are widely used in modern industrial manufacturing.Every day the world produces a large number of magnetic parts.Defects on the surface of the magnetic material will greatly reduce the performance of the magnetic material.The detection of apparent defects in magnetic materials is an important method of ensuring the quality of magnetic materials.At present the artificial visual inspection methods adopted by domestic manufacturers have greatly limited the development of the industry.The automatic optical inspection equipment adopted by a few manufacturers is difficult to meet the needs of actual production because of low detection efficiency.It is necessary to conduct further research on the surface detection of magnetic materials.Aiming at the automatic optical detection of apparent defects of magnetic materials,this paper presents a set of algorithm for detecting apparent defects of magnetic materials based on machine vision.The main contents of the study are:(1)Research on adaptive target area positioning technology.Aiming at the difficulty of segmentation and positioning caused by the differences in coating material coating thickness and coating process of magnetic materials this paper proposes an adaptive ergodic segmentation method to locate the target region.Firstly the adaptive ergodic method is used to perform image segmentation.By traversing the threshold an appropriate threshold interval is found.Finally the value of a certain point in the interval is the segmentation threshold.Then compare with the template information to determine whether the located area is the current batch of magnetic material.The method is compatible with magnetic materials of all materials has high robustness and has a better segmentation effect than a general segmentation algorithm.(2)Research on edge detection algorithm for magnetic materials.In this paper three methods are comprehensively compared.The first method is template matching.Using the template prepared in advance to subtract the image from the current workpiece to obtain the difference map.Experiments show that template matching can more easily detect missing edges and missing corners of magnetic materials.The second method is the pit search method.By using the method of searching for pits on the contour the large pit found is the side of the magnetic material.The third method is the convex hull detection method.By finding the convex hull of the outer contour the pattern enclosed by the outer contour and the contour hull of the outer contour are subtracted so that the defective portion is directly obtained.This paper also proposes an optimization method for the image subtraction part.The experimental results show that the algorithm of convex hull detection algorithm is stable and the ability to detect minor defects is stronger.This paper also analyzes the advantages and disadvantages of various methods and finally uses the method of combining template matching and convex hull algorithm to achieve high accuracy and efficient detection of edge defects.(3)Research on defect detection algorithm on magnetic material surface.This paper proposes a defect extraction method based on two-dimensional Gabor filter.By constructing multi-directional multi-scale Gabor filtering kernel and then comprehensively analyzing each filtering result mask processing and morphological opening operation are used to remove image edge and noise interference.Get defect information.This method solves the problem of difficult detection due to the complex background surface of magnetic images.(4)Acceleration and optimization of the algorithm.This article uses multithreading OpenMP and CUDA to accelerate the algorithm.Through experimental comparison CUDA acceleration is finally selected the detection time is reduced to 1/4 of the non-accelerated time and the time is controlled within 65 ms to meet the real-time detection requirements.After batch testing the missing detection rate and false detection rate of the edge detection algorithm in this paper decreased from about 10% to less than 1%.The rate of missed inspections and erroneous inspections on magnetic materials decreased from about 15% to less than 2%.The algorithm meets the requirements of practical applications.