The increasing prevalence of unmanned aerial vehicles (UAVs), commonly known as drones, has generated a demand for reliable detection systems. The inappropriate use of drones presents potential security and privacy hazards, particularly concerning sensitive facilities. Consequently, a critical necessity revolves around the development of a proficient system with the capability to precisely identify UAVs and other flying objects even in challenging scenarios. Although advancements have been made in deep learning models, obstacles such as computational intricacies, precision limitations, and scalability issues persist. To overcome those obstacles, we proposed the concept of MultiFeatureNet (MFNet), a solution that enhances feature representation by capturing the most concentrated feature maps. Additionally, we present MultiFeatureNet-Feature Attention (MFNet-FA), a technique that adaptively weights different channels of the input feature maps. To meet the requirements of multi-scale detection, we presented the versions of MFNet and MFNet-FA, namely the small (S), medium (M), and large (L). The outcomes reveal notable performance enhancements. For optimal bird detection, MFNet-M (Ablation study 2) achieves an impressive precision of 99.8%, while for UAV detection, MFNet-L (Ablation study 2) achieves a precision score of 97.2%. Among the options, MFNet-FA-S (Ablation study 3) emerges as the most resource-efficient alternative, considering its small feature map size, computational demands (GFLOPs), and operational efficiency (in frame per second). This makes it particularly suitable for deployment on hardware with limited capabilities. Additionally, MFNet-FA-S (Ablation study 3) stands out for its swift real-time inference and multiple-object detection due to the incorporation of the FA module. The proposed MFNet-L with the focus module (Ablation study 2) demonstrates the most remarkable classification outcomes, boasting an average precision of 98.4%, average recall of 96.6%, average mean average precision (mAP) of 98.3%, and average intersection over union (IoU) of 72.8%. To encourage reproducible research, the dataset, and code for MFNet are freely available as an open-source project: github.com/ZeeshanKaleem/MultiFeatureNet .

Unmanned aerial vehicles (UAVs) are popular in various applications due to their mobility, size, and user-friendliness. However, identifying malicious UAVs presents challenges that need to be encountered in general image-based object detection. These challenges arise because UAVs can fly at different altitudes, making it challenging to distinguish them from other flying objects and identify their size. In addition, the speed of UAVs also adds to the difficulty of capturing their clear images, which can lead to blurring, particularly in complex backgrounds. To address these challenges, we present an improved YOLOv5 architecture named M ulti- S cale F eature map G hostNet ( MSF-GhostNet ) by introducing GhostConv (GC) and C3Ghost (C3G) modules to reduce the redundant operations in the head and neck. We also proposed three feature map combinations to evaluate the performance of multi-scale and multi-target flying objects, including drones, birds, planes, and helicopters. This approach significantly reduces the waste of computing resources when detecting small-sized flying objects. We also integrated auto-anchor and batch size mechanisms to ensure efficient model training and avoid over-fitting. Our proposed model showed 1.25% fewer false positives (FP) than the state-of-the-art GhostNet-YOLOv5 model. The proposed MSF-GhostNet outperformed GhostNet-YOLOv5 with higher precision, recall, and F1 scores (1.3%, 5.3%, and 3.7% respectively) and reduced model parameters and model size by 3.1% and 4.1%, respectively. The proposed solution also outperformed several other state-of-the-art algorithms exists in the literature. We also provided dataset and code as an open-source: github.com/ZeeshanKaleem/MSF-GhostNet .

Vision sensors are versatile and can capture a wide range of visual cues, such as color, texture, shape, and depth. This versatility, along with the relatively inexpensive availability of machine vision cameras, played an important role in adopting vision-based environment perception systems in autonomous vehicles (AVs). However, vision-based perception systems can be easily affected by glare in the presence of a bright source of light, such as the sun or the headlights of the oncoming vehicle at night or simply by light reflecting off snow or ice-covered surfaces; scenarios encountered frequently during driving. In this paper, we investigate various glare reduction techniques, including the proposed saturated pixel-aware glare reduction technique for improved performance of the computer vision (CV) tasks employed by the perception layer of AVs. We evaluate these glare reduction methods based on various performance metrics of the CV algorithms used by the perception layer. Specifically, we considered object detection, object recognition, object tracking, depth estimation, and lane detection which are crucial for autonomous driving. The experimental findings validate the efficacy of the proposed glare reduction approach, showcasing enhanced performance across diverse perception tasks and remarkable resilience against varying levels of glare.

Robust lane detection is imperative for the realization of intelligent transportation. Recently, vision-based systems that employ deep convolution neural networks (CNNs) for lane detection have made considerable progress. However, for better generalization under various road conditions learning-based methods require excessive training data, which becomes non-trivial in challenging conditions such as illumination variation, shadows, false lane lines, and worn lane markings, etc. In this paper, we propose a light field (LF) based lane detection method that utilizes the additional angular information for improved prediction and increased robustness. Two different LF representations are investigated to study the possibility of maximum performance improvement and minimal additional computation cost and data labeling efforts. Experimental results successfully demonstrate that the proposed approach improves the prediction of the lane line point coordinates and is significantly robust against the aforementioned adverse conditions.

Lane detection is an essential module for the safe navigation of autonomous vehicles (AVs). Estimating the vehicle’s position and trajectory on the road is critical; however, several environmental variables can affect this task. State-of-the-art lane detection methods utilize convolutional neural networks (CNNs) as feature extractors to obtain relevant features through training using multiple kernel layers. It makes them vulnerable to any statistical change in the input data or noise affecting the spatial characteristics. In this paper, we compare six different CNN architectures to analyze the effect of various adverse conditions, including harsh weather, illumination variations, and shadows/occlusions, on lane detection. Among all the aforementioned adverse conditions, harsh weather in general and snowy night conditions particularly affect the performance by a large margin. The average detection accuracy of the networks decreased by 75.2%, and the root mean square error (RMSE) increased by 301.1%. Overall, the results show a noticeable drop in the networks’ accuracy for all adverse conditions because the features’ stochastic distributions change for each state.