This paper describes the development of a ground-mounted sensing system designed for urban air mobility (UAM) operational aircraft's landing and takeoff within designated UAM Vertiport zones. One of the significant challenges associated with adopting visible feature-based methods for detecting UAM operational aircraft is their complex shapes and patterns, which can complicate the extraction of consistent and distinctive visible features for accurate detection. To address this challenge, we developed a system that focuses on extracting keypoint features from foreground images, forming a basis for reliable region proposals encompassing all types of flying objects. The resulting system was experimentally evaluated using drones of varying shapes and sizes, flying at distances ranging from 15 to 120 meters. The experiments demonstrated the system's capability to detect small drones, as small as 0.178x0.232x0.127 meters, flying at approximately 100 meters. Additionally, the system exhibited the ability to detect flying birds. The paper also demonstrates the use of the developed visible image-based flying object region proposal system to assess the retraining and finetuning needs of state-of-the-art deep learning-based object detection models.