Research On Visual Inspection Method Based On Phase Congruency

Phase congruency model is an image enhancement method in frequency domain with phase information,and has the property of local contrast invariant and insensitive to illumination change.Therefore,it can express image features more stably and especially has unique advantages in dealing with low contrast image features,which is also an important reason why it is widely used in various image processing tasks.Considering the advantage of phase congruency model in characterizing image features,aiming at the problem that the evaluation indicators are mainly concentrated in spatial domain and the amplitude frequency domain for the surface roughness inspection based on machine vision,but the phase spectrum is ignored which has an advantage in expressing the structure of images,a method based on phase consistency model to detect the surface roughness is proposed,and aiming at the problem of being difficult in weld pool edge extraction in laser cladding due to the interference such as arc,molten metal drop and so on,A method for extracting the weld pool edge based on phase congruency is presented.The results of the related experiments show that the algorithm of grinding surface roughness evaluation based on phase congruency has high precision for grinding surface roughness detection and the method of weld pool edge extraction for laser cladding based on phase congruency model has good performance in the accuracy and robustness of weld pool edge extraction.The main research contents of this article are as follows:(1)By reviewing the research status about the application of phase congruency,surface roughness detection methods and the methods of weld pool edge extraction for laser cladding,summarize the reason why phase congruency model is widely used,and the problems of the algorithms of roughness evaluation and the methods of weld pool edge extraction in laser cladding are obtained.The content of the research is determined,and the research ideas are raveled out.(2)For the purpose of better applying the phase congruency model to solve the problems which are presented in the review,firstly,introduce the definition of phase congruency model,the calculation of phase congruency model and the construction of two dimensional phase congruency model,expound the mathematical definition and calculation model of various classical image edge algorithms,and the similarities and differences between the phase congruency model and the classical edge extraction algorithms are analyzed theoretically;Secondly,in order to understand the phase congruency model more intuitively,the performance of the algorithm related to phase information is detected by using some examples:The importance of phase information is emphasized by analyzing the function of phase information and amplitude information of the image;Compare the phase congruency model with the classical edge extraction algorithms,and the similarities and differences between them and their performance are demonstrated;The noise immunity of the edge extraction algorithm based on phase congruency is analyzed by simulation research of noisy image;The relations between the image,frequency and phase consistency model are studied from the frequency domain to further analyze and verify the applicable objects and advantages of the phase congruency model.Through the above research,the related performance and advantages of the phase congruency model are analyzed and validated from the theoretical aspect and the practical aspect,which is instructive for the solution of the problems presented in this paper.(3)Aiming at the problem that the evaluation indicators of frequency domain are mainly concentrated in the amplitude frequency spectrum for roughness inspection based on machine vision,but the phase spectrum frequency is ignored which has an advantage in expressing the structure of images,a new algorithm for the detection of grinding surface roughness based on phase congruency is presented.The method gets the difference charts by processing the virtual images of color swatch,the picture of phase congruency of color difference map is obtained by using phase congruency model,construct the image clarity feature index based on phase congruency model,and the model of relationship between image sharpness index and surface roughness is established by using SVM for the roughness detection.The experimental results show that the phase congruency model used in grinding surface roughness evaluation has higher detection accuracy.(4)Aiming at the problem of being difficult in weld pool edge extraction in laser cladding due to the interference such as arc,molten metal drop and so on.A method for extracting the weld pool edge based on phase congruency is proposed.The processing steps of the algorithm for weld pool image include weld pool image preprocessing,rough segmentation of molten pool,molten pool edge detection,further segmentation of weld pool and weld pool edge extraction.Compare it with other edge extraction algorithms qualitatively and quantitatively,and further verify that the algorithm has better robustness to the interference and edge extraction accuracy.The research work of this dissertation provides important technical support for studying the image characteristic index which has stronger relevance with grinding surface roughness and developing edge algorithm for laser cladding with stronger robustness to the interference,and has certain reference significance to the research of surface roughness inspection by machine vision and the research of machine vision based laser cladding quality real-time control.