Conclusion A simple imaging system was developed with a color camera and pair of illuminations. The developed computer vision system was tested on three popular date varieties for its performance. The mean classification accuracy was 68% to 86% in three classes model and 83% to 96% in two classes model for three varieties of dates.Although 39 features were extracted from the date images, around 10 features were sufficient to obtain similar classification results in both approaches. This method has great potential to be utilized for online quality monitoring of dates. However, the performance must be studied for bulk samples and samples in motion.