Research On Enhancement And Recognition Of Wood Veneer Knots Image Based On Adaptive Correction And USM

In the wood and wood processing industry,veneer is the species with the largest demand and an important part of wood-based panels such as plywood.The quality of veneer determines its use efficiency.Knot defect is the biggest factor affecting the quality and use of veneer.In production,most veneer quality inspection is still manual visual inspection,which is cumbersome and is very likely to have human errors.Therefore,combined with digital image processing technology and support vector machine pattern recognition technology,this thesis studies the live joint and dead joint on the veneer surface,including the preprocessing,feature extraction,recognition and classification of the veneer knot image.(1)Three classical image enhancement algorithms based on adaptive gamma correction and quantile,guided filtering,homomorphic filtering and Laplace sharpening are introduced.At the same time,these three algorithms are used as comparison algorithms to enhance the veneer knot image.It is found that while improving the contrast,it will cause the blur of detail areas such as the edge of knot defects.(2)From the perspective of contrast and detail enhancement,this thesis proposes a single board knot enhancement algorithm based on adaptive correction and unsharp masking(USM).The weighted distributed adaptive gamma correction algorithm,adaptive nonlinear stretching and USM are formed into a new enhancement algorithm.In this method,the brightness component and saturation component of HSV color space are extracted,and the brightness component is processed by weighted distribution adaptive gamma correction algorithm to enhance the contrast of veneer knot image.In order to avoid the unnatural color,the saturation component is adaptively nonlinear stretched.USM technology is used to process the veneer knot image after contrast enhancement to enhance its detail area.(3)According to the mean square error,peak signal-to-noise ratio,structural similarity,entropy,mean gradient,edge intensity,enhancement measure,and subjective evaluation,the quality of the enhanced veneer knot image is quantitatively and qualitatively evaluated.Among the four methods,the adaptive correction and USM algorithm proposed in this thesis have better performance in noise control,detail enhancement and contrast enhancement,which maximally improve the quality of segmenting veneer knot images.(4)This thesis combines the veneer knot enhancement algorithm based on adaptive correction and USM with support vector machine,describes the veneer knot defects by using the gray level co-occurrence matrix of texture features and the directional gradient histogram of local features,inputs the appropriate feature set for support vector machine,and establishes a complete set of methods suitable for the identification and classification of live and dead knots in veneer.The experimental results show that the proposed algorithm based on adaptive correction and USM sharpening shows good enhancement effect and applicability in the contrast and detail enhancement of veneer knot image,and the overall performance is better than the other three comparison algorithms;After preprocessing based on adaptive correction and USM sharpening algorithm,the recognition accuracy of veneer knot image has been greatly improved compared with the other three comparison algorithms.