Early Detection System of Cataract using Haar-like Feature-Based Cascade Classifiers

A Cataract is a disease that resists light entering the eye due to the cloud of the eye's lens. A cataract is the highest proportion in the cause of blindness in Indonesia which is 77.7%. Even so, cataracts blindness can be avoided with surgery. However, cataract diagnosis needs expensive tools and is not easy to use, for instance, slit lamp and ophthalmoscope. Therefore, a system that can be publicly used to detect cataracts easily is required. The cataract detection system is expected to increase the number of cataract surgery, so the number of blindness can be decreased. Digital image classification is an approach to tackle these issues. In this study, the system uses a grayscale image of the eye's lens which is extracted by Haar-like feature, then its class is predicted by the Cascade Classifiers method. Based on the K-Fold Cross Validation test, where k is 5, it is known that the average accuracy obtained is 82% and the highest accuracy obtained is 95%. It shows that the Haar Cascade Classifiers method can be used to classify cataracts to reduce the number of blind people. But, it is recommended to use a greater amount of training data to achieve a better result.