Browsing by Author "Yildirim, Suleyman"

Now showing 1 - 6 of 6
  • Thumbnail Image
    ItemOpen Access
    Autonomous ground refuelling approach for civil aircrafts using computer vision and robotics
    (IEEE, 2021-11-15) Yildirim, Suleyman; Rana, Zeeshan; Tang, Gilbert
    3D visual servoing systems need to detect the object and its pose in order to perform. As a result accurate, fast object detection and pose estimation play a vital role. Most visual servoing methods use low-level object detection and pose estimation algorithms. However, many approaches detect objects in 2D RGB sequences for servoing, which lacks reliability when estimating the object’s pose in 3D space. To cope with these problems, firstly, a joint feature extractor is employed to fuse the object’s 2D RGB image and 3D point cloud data. At this point, a novel method called PosEst is proposed to exploit the correlation between 2D and 3D features. Here are the results of the custom model using test data; precision: 0,9756, recall: 0.9876, F1 Score(beta=1): 0.9815, F1 Score(beta=2): 0.9779. The method used in this study can be easily implemented to 3D grasping and 3D tracking problems to make the solutions faster and more accurate. In a period where electric vehicles and autonomous systems are gradually becoming a part of our lives, this study offers a safer, more efficient and more comfortable environment.
  • No Thumbnail Available
    ItemOpen Access
    Data supporting: 'Autonomous Ground Refuelling Approach for Civil Aircrafts using Computer Vision and Robotics'
    (Cranfield University, 2022-08-15 09:05) Yildirim, Suleyman
    Aircraft Refuelling Adaptor Localisation - v3 2022-03-15 12:35pm It includes 881 images. Letters are annotated in COCO format. The following pre-processing was applied to each image: The following augmentation was applied to create 3 versions of each source image: * 50% probability of horizontal flip * 50% probability of vertical flip * Equal probability of one of the following 90-degree rotations: none, clockwise, counter-clockwise, upside-down * Random Gaussian blur of between 0 and 1 pixels
  • Thumbnail Image
    ItemOpen Access
    Development of vision guided real-time trajectory planning system for autonomous ground refuelling operations using hybrid dataset
    (AIAA, 2023-01-19) Yildirim, Suleyman; Rana, Zeeshan A.; Tang, Gibert
    Accurate and rapid object localisation and pose estimation are playing key roles during some of the real-time robotic operations such as object grasping and object manipulating. To do so, high-level robotic vision solutions need to be adopted. Computer vision approaches require a large amount of data to be able to create a perception pipeline robustly. Preparing such dataset to train the deep neural network could be challenging as the collection and manual annotation of huge amounts of data can take long hours and the development of the dataset needs to cover different conditions in weather and lighting. To ease this process, generating a synthetic dataset could be used. Due to the limitations of the synthetic dataset which will be described further down, instead of using a sole synthetic dataset, a hybrid dataset can be developed with the real dataset to overcome the limitations of both datasets. Even though the main objective of this study is to fulfil an autonomous nozzle insertion process for the ground refuelling operation of civil aircraft, the proposed approach is generic and can be adapted to any 3D visual robotic manipulation operation. This study is also offered to be the first visual trajectory planning control mechanism depending on the hybrid dataset to this date.
  • Thumbnail Image
    ItemOpen Access
    Enhancing aircraft safety through advanced engine health monitoring with long short-term memory
    (MDPI, 2024-01-14) Yildirim, Suleyman; Rana, Zeeshan A.
    Predictive maintenance holds a crucial role in various industries such as the automotive, aviation and factory automation industries when it comes to expensive engine upkeep. Predicting engine maintenance intervals is vital for devising effective business management strategies, enhancing occupational safety and optimising efficiency. To achieve predictive maintenance, engine sensor data are harnessed to assess the wear and tear of engines. In this research, a Long Short-Term Memory (LSTM) architecture was employed to forecast the remaining lifespan of aircraft engines. The LSTM model was evaluated using the NASA Turbofan Engine Corruption Simulation dataset and its performance was benchmarked against alternative methodologies. The results of these applications demonstrated exceptional outcomes, with the LSTM model achieving the highest classification accuracy at 98.916% and the lowest mean average absolute error at 1.284%.
  • Thumbnail Image
    ItemOpen Access
    The influence of micro-expressions on deception detection
    (Springer, 2023-03-16) Yildirim, Suleyman; Chimeumanu, Meshack Sandra; Rana, Zeeshan A.
    Facial micro-expressions are universal symbols of emotions that provide cohesion to interpersonal communication. At the same time, the changes in micro-expressions are considered to be the most important hints in the psychology of emotion. Furthermore, analysis and recognition of these micro-expressions have pervaded in various areas such as security and psychology. In security-related matters, micro-expressions are widely used to detect deception. In this research, a deep learning model that interprets the changes in the face into meaningful information has been trained using The Facial Expression Recognition 2013 dataset. Necessary data is also obtained through live stream or video stream by detecting via computer vision and evaluating with the trained model. Finally, the data obtained is transformed into graphic and interpreted to determine whether the people are trying to deceive or not. The deception classification accuracy of the custom trained model is 74.17% and the detection of the face with high precision using the computer vision methods increased the accuracy of the obtained data and provided it to be interpreted correctly. In this respect, the study differs from other studies using the same dataset. In addition, it is aimed to facilitate the deception detection which is performed in a complex and expensive way, by making it simple and understandable.
  • Thumbnail Image
    ItemOpen Access
    Reducing the reality gap using hybrid data for real-time autonomous operations
    (MDPI, 2023-04-02) Yildirim, Suleyman; Rana, Zeeshan
    This paper presents an ablation study aimed at investigating the impact of a hybrid dataset, domain randomisation, and custom-designed neural network architecture on the performance of object localisation. In this regard, real images were gathered from the Boeing 737-400 aircraft while synthetic images were generated using the domain randomisation technique involved randomising various parameters of the simulation environment in a photo-realistic manner. The study results indicated that the use of the hybrid dataset, domain randomisation, and the custom-designed neural network architecture yielded a significant enhancement in object localisation performance. Furthermore, the study demonstrated that domain randomisation facilitated the reduction of the reality gap between the real-world and simulation environments, leading to a better generalisation of the neural network architecture on real-world data. Additionally, the ablation study delved into the impact of each randomisation parameter on the neural network architecture’s performance. The insights gleaned from this investigation shed light on the importance of each constituent component of the proposed methodology and how they interact to enhance object localisation performance. The study affirms that deploying a hybrid dataset, domain randomisation, and custom-designed neural network architecture is an effective approach to training deep neural networks for object localisation tasks. The findings of this study can be applied to a wide range of computer vision applications, particularly in scenarios where collecting large amounts of labelled real-world data is challenging. The study employed a custom-designed neural network architecture that achieved 99.19% accuracy, 98.26% precision, 99.58% recall, and 97.92% mAP@.95 trained using a hybrid dataset comprising synthetic and real images.