PCA Preprocessing Target Spectrogram for Birds and Drones Classification

Abstract

As the growing of the consumer market of personal drones, technologies are needed to monitor drones in the airspace. Radar has been proven to be a suitable tool for the detection and surveillance of the drone. One of the main obstacles in drone detection is the interference of the birds; since two targets share the same flight level, ground speed, and size. In this Final Year Project research, the impact of principal component analysis(PCA) on the classification of birds and drones is investigated and demonstrated. The whole project was built and tested on the L band staring radar dataset. Several different algorithm of PCA is implemented first and a classifier is trained based on support vector machine(SVM). The SVM classifier is used to demonstrate the impact of PCA preprocessing on classification. The impact of preprocessing is evaluated by the results such as training time cost, confusion matrix and general accuracy of classification. Finally, the different performance of PCA preprocessing algorithm is demonstrated and the reason of these result is concluded. It shows that PCA preprocessing can be an efficient way to increase the performance of the classifier when facing the radar spectrogram data.