Deep Learning Based Vehicle Detection on Real and Synthetic Aerial Images: Training Data Composition and Statistical Influence Analysis

Abstract

The performance of deep learning based algorithms is significantly influenced by the quantity and quality of the available training and test datasets. Since data acquisition is complex and expensive, especially in the field of airborne sensor data evaluation, the use of virtual simulation environments for generating synthetic data are increasingly sought. In this article, the complete process chain is evaluated regarding the use of synthetic data based on vehicle detection. Among other things, content-equivalent real and synthetic aerial images are used in the process. This includes, in the first step, the learning of models with different training data configurations and the evaluation of the resulting detection performance. Subsequently, a statistical evaluation procedure based on a classification chain with image descriptors as features is used to identify important influencing factors in this respect. The resulting findings are finally incorporated into the synthetic training data generation and in the last step, it is investigated to what extent an increase of the detection performance is possible. The overall objective of the experiments is to derive design guidelines for the generation and use of synthetic data.

Automated Recognition and Tracking of Stationary and Moving Cars in Images and Videos: YOLOv5 and SSD Analysis

“…Traditional tracking methods work by first identifying items in the initial frames, and then searching the surrounding environment for features that correspond to those things to locate a picture sequence that best fits them. Traditional detectors, such as contour-based target tracking [34] and the Harris corner detection [35], produced false detection, SIFT (symmetric integral and fluctuating transform), and point-based approaches [36,37]. But, by first detecting the objects with DL models, and then proceeding to match features via the conventional tracking methodologies, better performance was attained.…”

“…To ensure that certain feature points are not included in the final verdict, we employ the motion data that is linked to them. Clustering strategies are employed to organize the feature points into groups, and some are detailed in [12,36,37]. However, K-means clustering is sufficient for the needs of this task, as it yields a high degree of accuracy while keeping the computational complexity under control.…”