Modern navigation systems face significant challenges due to the limitations of traditional methods such as Inertial Navigation Systems (INS) and Global Navigation Satellite System (GNSS). INS suffers from cumulative drift errors, while GPS is vulnerable to jamming, spoofing, and signal loss, particularly in adversarial or obstructed environments. Digital Scene Matching and Area Correlation (DSMAC) is one of effective method used to alleviate these issues. Traditionally, DSMAC is implemented using image processing techniques. In this paper, we introduce ML based solutions which are easy and fast to build and trained for a specific mission. A custom dataset of 35,000 aerial images was created based on the planned flight path and altitudes, to train and evaluate various Convolutional Neural Network (CNN)-based models, including Facebook AI Similarity Search(FAISS) and traditional image matching model i.e. Oriented FAST and rotated BRIEF (ORB). These models, originally designed for tasks such as image segmentation and image classification, were fine-tuned specifically for aerial image matching. Through this fine-tuning process and subsequent testing, performance metrics were obtained, demonstrating the potential of these models to handle the unique requirements of aerial navigation such as, handling images captured from various angles, image distortion due to weather conditions and real-time processing with limited compute resources. Additionally, a data creation application was developed around Google Earth Pro, which generates datasets based on marked areas provided in a KML file. This application automates dataset creation by extracting aerial images of specified regions, automating the data preparation process.