Wood failure evaluation is the key criterion for predicting the long-term durability of plywood. At present, the conventional visual method for plywood wood failure evaluation is slow and subjective. Even experienced evaluators can show significant differences in their evaluations on the same plywood specimen and an individual evaluator can make different wood failure estimates on the same specimen at different times. Differences among evaluators can be as high as 50% for some samples. Evaluations can be influenced by room lighting, the wood species, sample treatment, and readings from prior samples. An automatic wood failure evaluation system using image analysis techniques could potentially be programmed to consider all the variables and respond with consistent wood failure values regardless of the experience level of the machine operator. This report describes the results of a one-year project in which a system for automatic plywood wood failure determination was investigated. A color optical imaging system was assembled and the preliminary work of compiling suitable algorithms was completed with promising results. The imaging system was 100% effective in reproducing individual sample values. Samples were sorted according to plywood type and test method to develop appropriate program algorithms for each category. The wood failure program was then further developed to automatically detect wood species and test method, thus avoiding the need for specimen separation prior to evaluation. Based on nearly 1200 samples in four categories, the differences in average values of wood failure between human evaluation and machine vision were found to be less than plus or minus 5%. In addition, a minimum of 85% of individual machine readings fell in the plus or minus 15% range of deviation expected of human wood failure readers. The imaging system was more accurate for light-colored specimens (i.e., Canadian Softwood Plywood) than darker-colored specimens (i.e., Douglas fir ) and for specimens where resin had been applied by spray. In order to make the imaging system more reliable and robust, the algorithm parameters now need to be fine-tuned based on a larger sample database.