Research On Periodic Structure Detection Method Based On Spectrum Characteristics Analysis

Periodic structure is a common feature,widely used in many fields ranging from buildings,bridges,machinery and fabrics to cells,compound eyes,chips and microlens arrays.In order to ensure the consistency and accuracy of the structure,the detection of periodic structural parameters is indispensable.With the development of image processing technology and computer vision technology,more and more measurement can be automatically completed by visual system,which improves the accuracy and speed of detection,avoids the possible errors in manual detection,and also improves the efficiency and reduces the cost.In this paper,the frequency-domain detection method is systematically and comprehensively analyzed and studied in many periodic structural parameter detection methods.The main work of this paper is as follows:1)A spectrum feature extraction method is proposed,and the influence of imaging errors such as noise,shadow and distortion on the spectrum and the influence of image acquisition errors such as tilt and zoom on the spectrum are analyzed.And then,a preprocessing method is proposed.2)In order to meet the needs of dimension measurement,quality detection and structural similarity evaluation of periodic structures,texture image simulation is carried out by using control variables to extract spectrum features and find out the corresponding spectrum features of each texture to be measured.Through quantitative analysis and calculation,an evaluation index and a general detection method based on spectral feature information are proposed for each texture to be measured.3)Visual inspection system and user interface are built,and the actual periodic structure represented by knitted fabrics is experimented.The detection methods based on spectrum characteristics,such as dimension measurement,quality detection and structural similarity evaluation are verified.4)A real-time acquisition and detection system have been built and successfully applied to real-time fabric density detection.Through the experiment and analysis of different samples,the measuring speed of the system is about 150 milliseconds,and the measuring error is about 1%.This shows that the detection algorithm of periodic structure parameters based on spectrum analysis is accurate,efficient and universal.