A fast detection method for building wood measurement based on interval type-2 fuzzy model

Abstract

Automatic measurement detection in the process of wood processing can effectively improve production efficiency and yield to promote upgrading the wood processing industry. Therefore, this article researches visual detection technology for wood surface measurements and proposes a construction wood measurement detection method based on interval type-2 fuzzy theory. This method overcomes the problem of increasing heterogeneity and similarity of homogeneous areas on the wood surface, which increases the difficulty of measurement detection. Firstly, the Gaussian model is improved by establishing a wood measurement image feature model that considers rate neighborhood relationships as the basic model. Then, the basic model is fuzzified to construct an interval type-2 fuzzy model (IT2FM) to characterize the uncertainty features of the modeling and achieve measurement detection of building wood. We conducted experiments using a wood measurement detection dataset to study the influence of the construction form of IT2FM on wood measurement detection and evaluated the results using kappa values. The detection accuracy of this method is improved by about 10% compared to traditional machine learning and deep learning methods, and it only takes about 3–5 s to run the program. The experimental results show that the proposed method can effectively suppress the noise generated by wood measurement detection and prevent missed detection. It can achieve high-quality measurement detection of wood dead knots, loose knots, and cracked parts.