Detection of Aircraft in Satellite Images using Multilayer Convolution Neural Network

The automatic identification of various spatial entities within satellite imagery is a crucial undertaking for interpreting such images. Numerous research papers have explored the segregation, identification, and geolocation of objects of interest, such as airplanes, vehicles, and human elements, within satellite images. The detection of aircraft from satellite imagery is particularly significant for gathering operational intelligence. Detecting aircraft within the environment is achieved through active remote sensing methods, such as RADAR and LASER. Various algorithms have been devised and developed specifically for aircraft detection in satellite imagery. The advent of AI-based techniques has brought about a transformative shift in object detection within remotely sensed images. This paper proposes a methodology employing a convolutional neural network (CNN) for the detection of aircraft within satellite imagery. Initially, an image dataset is generated using QGIS software, which is then partitioned into training and testing datasets. A multi-layered CNN model is employed to train and evaluate the dataset. Subsequently, the trained CNN is applied to remotely sensed images to detect the presence of aircraft within the scene. The aircraft detection accuracy from randomly selected satellite images is reported to be 95%.