Automated Extraction of Objects from Aerial Images using A Convolutional Neural Network

The amount of aerial images collected by drones and satellites increasing at an unprecedented rate, sophisticated techniques for autonomous object detection are required. In this research study, a MATLAB-implemented convolutional neural network (CNN) approach for automatic object detection in aerial photos is presented. The suggested system is made to precisely identify and classify objects within aerial images by utilising CNNs' potent feature extraction capabilities. This is important for a number of applications, such as surveillance, monitoring agriculture, and urban planning. The suggested method entails creating a deep learning model that is fed a substantial collection of annotated aerial photos for training. Using techniques for data augmentation to enhance model robustness and generalisation, as well as the integration of MATLAB's Deep Learning Toolbox for model construction and evaluation, are important breakthroughs. Extensive experiments were conducted to validate the CNN architecture's performance across various scenarios and image types. The results showed that the CNN-based system achieves significant improvements in detection accuracy and speed when compared to traditional image processing methods. The CNN architecture is optimised for high accuracy and computational efficiency. Comparative study using the most recent object detection frameworks confirms the approach's efficacy even more. This work offers an efficient and scalable method for autonomous object recognition in aerial photos, which could have ramifications for numerous practical uses.