Browsing by Author "Patel Amir"

Now showing 1 - 2 of 2
  • No Thumbnail Available
    Item
    Open Access
    Lightweight Mapping of Unstructured Environments
    (2024) Katsoulis, Michael; Amayo, Paul; Patel Amir
    We present a computationally, size and cost-lightweight monocular reconstruction pipeline that produces high-quality reconstructions in an unstructured agricultural environment consisting of orchards and vineyards. The pipeline has to be deployed and tested on ground vehicles equipped only with a single monocular camera sensor. Running on a CPU only, while achieving a sufficient resolution to identify individual plants without limitations on maximum path length. We show that a simple visual odometry system is capable of providing performance that is more accurate than GPS, with a relative transform error of 0.19m, without the need for techniques such as bundle adjustment or loop closure. Towards this contribution, we evaluate the impact of the choice of image feature on the accuracy of the visual odometry as well as the impact of the choice of disparity estimation method on the accuracy. Additionally, we show that state-of-the-art unsupervised monocular depth networks can outperform stereo techniques in terms of accuracy achieved when estimating the depth in an agricultural setting. We also show that lightweight pyramid-based methods are able to match the performance of deep monocular depth networks at the task of disparity estimation. The pipeline presented is optimised for application in agricultural environments and is lightweight in terms of size, weight, power and computational requirements. The pipeline functions using only a single camera and without any other sensors or sources of additional information making
  • No Thumbnail Available
    Item
    Open Access
    Platform and Pipeline Development for a FMCW Radar System for Vital Sign Detection
    (2025) Bowden, Nicholas; Paine, Stephen; Patel Amir
    Frequency Modulated Continuous Wave (FMCW) Radar has become a frequently examined technology for the purposes of ubiquitous vital sign monitoring applications. Vital sign monitoring of heart rate and respiration rate is important because it gives key physiological insights into the health of individual people. Having vital sign technology that allows for constant and ubiquitous monitoring would be of great benefit to people and physicians around the world. Most vital sign research outputs focus on singular parts of the vital sign monitoring problem, often heavily relying on machine learning and enormous datasets of pre-processed data to detect vital signs and compensate for artefacts introduced by breathing and other motion. This dissertation details the design of a system from the ground up to collect raw and unprocessed data and then goes further to explain the design a processing pipeline to validate the data from the system. This provided maximum versatility and flexibility for future research outputs. To validate the system and pipeline for vital sign detection, several sets of experiments were done with increasing complexity to identify points of failure within the pipeline. Complexity was added by adding layers of motion. First, the participant was in seated position and recordings were taken while the participant held his breath. Second, again in a seated position, recordings were taken while the participant was asked to inhale and exhale to visual cues. For these sets of experiments the pipeline performed well with accuracy ranging from 80% to over 90%. For the third set of experiments, the participant was asked to walk backwards and forwards during the recording session. Even after compensating for the movement, the accuracy of the system dropped significantly to below 60%. Compensation for large scale motion was achieved using a simple test rig by subtracting the known motion from the signal. However, the human walking motion was too complex to remove with just a simple subtraction. This complexity comes from the fact that walking requires multiple moving parts and these are all measured by the radar whereas with the rig, there is only one part that is moving. After exhausting traditional filtering and other standard Digital Signal Processing (DSP) techniques, this dissertation concludes that future work should probably adopt a Machine Learning (ML) approach to compensate for complex motions such as walking.