Automatic non-destructive quality inspection system for oil palm fruits

In this research a non-destructive, rapid and cost effective examination machine for the estimation of the ripeness fraction, oil content and free fatty acid level in oil palm fresh fruits bunch was developed. The automatic machine-vision based in- spection system provided consistency, rapid estimation and accep- table accuracy results in non-dest ructive manner. Fresh fruits bunch samples from Tenera cultivar (7 to 20 years trees) were taken from Cimulang plantation, Bogor, Indonesia. Two statistical analysis methods were used: a forward stepwise multiple linear regression analysis and a multilayer-perceptron artificial neural network analysis. The best prediction of ripeness and oil content models were obtained using the latter method, while the best free fatty acid prediction model was developed by the first method. The models were then employed in the machine-vision inspection systems of the machine. The system best prediction accuracy of ripeness, oil content and free fatty acid models was 93.5, 96.41, and 89.32%, with standard error of prediction being 0.065, 0.044 and 0.068, respectively. The system was tested through a series of field tests, and successfully examined more than 12 t of fruits bunch per hour, without causing damage.