James W. Funck

Image segmentation algorithms applied to wood defect detection

Abstract Image segmentation is a key stage in the detection of defects in images of wood surfaces. While there are many segmentation algorithms, they can be broadly divided into two categories based on whether they use discontinuities or similarities in the image data. Each algorithm can also be categorized based on other factors such as whether it uses color or gray-scale data and is a local or global operator. While this presents a wide variety of approaches for segmenting images of features on wood surfaces, it also makes it difficult to select the most appropriate techniques. This paper presents the results obtained from using a variety of algorithms for wood surface feature detection and defines several measures used for examining algorithm performance. A region-based, similarity algorithm that was a combination of clustering and region-growing techniques exhibited the best overall performance. This was particularly true for defects that are subtle, meaning they blend in with other natural features on wood surfaces that are not considered defects. Examples include blue stain, pitch streaks, and wane. The clustering with region growing algorithm improved the detection accuracy of pitch streaks by over 20 percentage points compared to the next best algorithm. However, if subtle defects are not of interest, the edge detection algorithms performed as well as the region growing algorithm but with slightly better clearwood detection accuracies. The influence of color information, local-basis analysis, and camera resolution on algorithm performance varied by segmentation technique and defect category. Because each wood processing application has its own unique set of defect detection requirements, conclusions regarding which algorithms and factors are best must be made in the context of those processing requirements.

A Discrete-Event Object-Oriented Modeling Environment For Sawmill Simulation

The paper presents a simulation modeling environment developed for sawmill design and analysis in the forest products industry. The design facilitates flexibility in modeling different sawmill configurations and production scenarios. The system represents a library of objects developed in an object-oriented framework. These include structures required to develop simulation models, to execute discrete event simulation, and for system-user interface.

Measuring surface roughness on wood: a comparison of laser-scatter and stylus-tracing approaches

Many wood products manufacturing processes require a 3-dimensional measure of surface roughness to determine processing parameters and product grades and values. Currently, on- line measurement of wood surface roughness is limited to visual inspection and single-point laser-based triangulation or ultrasonic systems, while most off-line analysis is based on stylus tracing. Wood has unique characteristics that complicate surface texture measurement and analysis such as the need to separate distinct causes of error of form, waviness, and roughness as well as to correlate visual grades of processing standards with 1-dimensional (1-D), 2-D, and 3-D measures. This paper discusses the performance characteristics of a laser scatter/optical imaging system for wood roughness measurement and compares them to those of a stylus tracing system. The abilities of both approaches to capture the types of roughness information required in wood manufacturing processes are discussed as well as the functionality of 1-D, 2-D, and 3-D roughness descriptors.

An evaluation of color spaces fordetecting defects in Douglas-fir veneer

Abstract Color information can be valuable for detecting and classifying surface defects inwood, but its usefulness depends on the color datas format and the analysis technique used. This study investigates five color transforms that convert the National Television Standards Committee red, green, and blue (RGB) primary color space into other potentially more useful spaces. A quadratic classifier was used to evaluate the relative utility of the different color spaces in separating defects from clear wood. Images of Douglas-fir veneer with encased knots, intergrown knots, and pitch streaks were converted to the various color spaces and then analyzed. The results show, for the conditions in this study, that a two-dimensional feature space is sufficient for classification and that there are no practically important differences in performance among the different color spaces. Thus, for images of Douglas-fir veneer, it appears that there is no advantage in mathematically transforming the original RGB data into another color space.

Wood-Surface Feature Classification Using Extended-Color Information

Pseudotsuga menziesii (Mirb.) Franco) veneer can be accurately classified via visible-region spectral-reflectance curves and quadratic discriminant analysis. This paper extends those results to an expanded set of features, a broader spectrum that includes near-infrared as well as visible wavelengths and a larger set of physical samples. It also tests two methods for eliminating the classification procedures reliance on raw spectral-reflectance curves. Instead of working with the raw curves, which are difficult to obtain by traditional means at sufficient speeds in a production environment, the data required by these two methods is much reduced and can potentially be obtained from a video camera equipped with either custom or commercially available bandpass filters. The paper shows that classification accuracies achieved with either of the two reduced-data methods are comparable to the accuracies achieved when using raw spectral data.Pseudotsuga menziesii (Mirb.) Franco) zuverlässig klassifiziert werden kann aufgrund der Daten aus Reflektionsspektren im sichtbaren Bereich und einer quadratischen Dirkriminatenanalyse. In diesem Beitrag werden die Ergebnisse erweitert hinsichtlich eines größeren Merkmalkatalogs, eines breiteren Spektralbereichs, der ausser dem sichtbaren auch den nahen IR-Bereich umfaßt, sowie eines größeren Probenumfangs. Weiter werden zwei Methoden geprüft, um die Abhängigkeit der Klassifikationsroutine von den spektralen Rohdaten zu eliminieren. Die Rohdaten-Kurven sind schwierig zu gewinnen mit konventionellen Methoden und auch nicht genügen rasch bei realen Produktionsbedingungen. Statt dessen werden die benötigten Daten stark reduziert. Diese können mit Hilfe einer Videokamera gewonnen werden, die mit einem eigenen oder kommerziellen Bandpassfilter ausgerüstet ist. Mit beiden Methoden ist eine Klassifikation aufgrund der reduzierten Daten mit vergleichbarer Genauigkeit erreichbar wie unter Nutzung der Originalspektren.

Classifying wood-surface features using dichromatic reflection

The dichromatic reflection paradigm describes light reflection from optically inhomogeneous materials as the sum of body (diffuse) and interface (specular) reflections. Interface reflection represents unaltered light reflected from a materials surface. Body reflection represents light altered by the materials pigments and thus may provide information about the identity of the material. Wood is an optically inhomogeneous material that is also anisotropic. This latter property adds further complexity to the analysis of wood-surface images by creating localized magnitude differences in interface reflection as surface texture and fiber orientation change. This paper presents the results of a study that tested whether the use of only the body component of reflected light can significantly improve the classification of wood-surface features. To this end, reflectance curves of various Douglas-fir (Pseudotsuga menziesii (Mirb.) Franco) veneer features were separated into body and interface components, and their dimensionality reduced to a small number of basis function. Two discriminant functions, one constructed from body reflectances and the other from total reflectances, were then developed from the reduced reflectance data. The performance of the two discriminant functions were compared by classifying a new set of wood-feature spectral reflectances with each discriminant function.

A note on face veneer checking in maple plywood

Abstract A common challenge related to hardwood plywood is checking along the grain of face veneers. This study tested the hypothesis that face checking of rotary peeled maple veneer plywood will be greater on the loose side of the veneer and that checking will be reduced by the application of a clear film-forming finish. Fifteen panels were constructed with the tight side of the face veneer oriented out (i.e. exposed to the atmosphere) and 15 with the loose side of the face veneer oriented out. Half of each panel was finished with lacquer and the other half left unfinished. Panels were conditioned in a hot, wet chamber, visually inspected, and then inspected again after conditioning in a hot, dry chamber. Contrary to conventional wisdom, panels oriented with the tight side of the face veneer oriented out checked significantly more than panels with the loose side oriented out. There was no effect due to finishing. Because these findings are contrary to established practices, further investigation is warranted.

Using simulation to select part-families for cell manufacturing

Abstract This paper examines how 11 part-families were developed and then selected, using discrete event simulation, for cell manufacturing and subsequent exclusion from the current manufacturing system of a value added wood products company. The criteria used to exclude a family were the reduction of current average work-in-process and, secondarily, average lead times. The excluded parts could then be processed through a manufacturing cell. Using simulation and management input, two of 11 part-families, representing 51 different part types, were chosen to be excluded from the traditional job shop floor. Their exclusion resulted in an average total work-in-process reduction of 112 parts which represented a 17% reduction. A second article, in preparation, will compare the processing of the 51 parts through different simulated manufacturing cell designs.

An expert system for softwood lumber grading

Abstract To accommodate future applications of scanning technology and minimize differences in the interpretation of written rules, a prototype expert system for grading green softwood lumber was developed. It was also intended to be part of the expansion of a log breakdown model to include internal defect information in the optimization process. The systems knowledge base consists of pertinent green grading rules based on the Western Wood Products Associations (1988) “Western Lumber Grading Rules 88” for 27 grades in the dimension, selects and finish, and board categories. The system is designed to be either interactive and menu-driven or run in a batch input mode. User input to the system consists of lumber size; desired primary grade category; and defect information such as type, location, and size on each face and edge. The system then infers the grade corresponding to each side and an overall grade for the board. Limited explanation capabilities are provided.

An object-oriented modeling framework for sawmill simulation

Abstract This paper describes a flexible, general-purpose simulation environment for sawmill modeling and analysis. The system represents a library of objects developed in an object-oriented framework. The objective is to develop a system that can be used to mode different sawmill configurations, to identify manufacturing process constraints, and to evaluate control strategies and management practices.