The best way to conference proceedings by Francis Academic Press

Web of Proceedings - Francis Academic Press
Web of Proceedings - Francis Academic Press

Fast Adaptive Machine Vision Localization Algorithm Based on Support Vector Regression Prediction

Download as PDF

DOI: 10.25236/icmmct.2023.025

Author(s)

He Huang, Honghu Ge

Corresponding Author

He Huang

Abstract

In this paper, a fast adaptive machine vision positioning algorithm based on support vector regression prediction is proposed to solve the problems of low efficiency and slow calculation speed of traditional machine vision positioning algorithms. Firstly, the algorithm uses relative threshold binarization to the template and the sampled image, which not only effectively overcomes the influence of illumination, but also reduces the amount of data. Then, it uses the Golden Tower of the two-layer image to reduce the calculation amount of the algorithm. In the matching search, it uses the method of adaptive step length to further speed up the matching, and can set the termination conditions to further speed up the speed. The algorithm quickly locates near the target by rough matching, and then locates at the center of the target by fine matching, which greatly improves the matching speed and ensures the accuracy of the algorithm. The algorithm is fast and accurate, and can meet the real-time requirements. In this paper, the method of applying support vector machine (SVM) to air traffic flow forecast is studied, an autoregressive forecast model based on SV M is established, and some key problems such as the determination of model parameters are discussed. On the whole, the prediction accuracy and stability of the combined prediction model are better than SVM prediction model. Support vector machine can solve the traditional difficulties encountered in the research of machine learning, such as over-learning and under-learning, local minima and small samples.In many applications of machine vision, especially in the field of semiconductor industry manufacturing, it is often necessary to perform efficient target recognition and matching for specific objects, which requires the image matching algorithm to have high robustness, and also requires less matching time and Higher matching accuracy. By providing the system with prototypes of such objects to be measured, it is very useful to find all types of objects in other images. Template matching is one such method. As an effective pattern recognition technique, it can more directly reflect the similarity between images by using image information and prior knowledge about the recognized pattern.

Keywords

Support vector machine; Regression prediction; Machine vision positioning algorithm; Prediction model