Image processing using sparse representation for classification with semi-random projection dimension reduction for the image recognition system

Abstract

Image recognition is a technology to identify objects, places, people and several other variables in digital images. The algorithm that can be used in the image recognition system is Sparse Representation for Classification. However, the high computational load is a problem in this study. In addition, there is a lot of training data needed to meet the sparse condition which is a weakness of this algorithm. Thus, to overcome this problem, dimensionality reduction can be carried out on the image using the Semi Random Projection method. In this study, the author used PyCharm software to process the Guide User Interface (GUI) of the dimensionality reduction system and image recognition using Semi Random Projection-Sparse Representation for Classification. Testing was carried out using 100 training data in the form of 50 red-green-blue images and 50 grayscale images. The images are divided into 10 classes and use 10 test images that have been added with noise and occlusion. Testing was carried out five times each on each test image. From the testing in this study, the results obtained on good performance parameters with an average accuracy of 98.93%, an average Peak Signal Noise to Ratio (PSNR) of 33.392741 dB and an average computing time of 1112.57952 ms.