Browsing by Author "Pickard, Joshua K."

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Deformation estimation of industrial objects from a single image
    (University of New Brunswick, 2024-09) Eivazi Adli, Sahand; Dubay, Rickey; Pickard, Joshua K.; Sun, Grace
    Deformations introduced during the manufacturing process of plastic components degrade the accuracy of their 3D geometric information, hindering computer vision-based inspection. This phenomenon is prevalent among the primary plastic products where the objects are devoid of texture. This work proposes a solution for the deformation estimation of texture-less plastic objects using only a single RGB image. This solution encompasses a unique image dataset of five deformed parts, including both real-world and synthetic images, a novel method for generating mesh labels, sequential deformation, and a training model based on graph convolution. The sequential deformation method overcomes the prevalent chamfer distance algorithm in generating precise mesh labels. The model achieves a sub-millimeter accuracy on synthetic images and approximately 2.0 mm on real images, with an average testing time of 1.5 s on the Google Colab’s resources. The model’s high precision and speed make it suitable for real-world applications.
  • Thumbnail Image
    Item
    Intelligent quality inspection methods for industrial applications
    (University of New Brunswick, 2024-04) Parrott, Edward Barrington Allaby; Dubay, Rickey; Pickard, Joshua K.
    Visual quality inspection of manufactured parts is a complex, non-linear problem whose methods are mostly based on human intuition and heuristics. However, the demand for parts with ever-increasing quality and precision has exposed the need for better, more intelligent inspection methods, which has led to the adoption of machine vision-based inspection in a wide range of manufacturing fields. While these methods have led to notable improvements in quality inspection, the performance of a machine vision-based inspection system is intrinsically linked to how effectively it is able to image the part that it must evaluate. As such, an ideal inspection system must be able to observe the part in its entirety using the fewest images possible. While work has been done in an attempt to solve this problem, most approaches rely heavily on heuristic global optimization methods which are inherently unable to guarantee the optimality of their solutions, and whose solutions are often lacking in interpretability while also being computationally expensive. Additionally, the quality of the results of these systems is often extremely sensitive to variations in part geometry and environmental conditions. This research aims to postulate the quality inspection problem using set-based approaches, which will allow the application of interval analysis tools to develop robust optimal inspection methods. The proposed methods aim to rigorously generate inspection spaces which are not only guaranteed to contain feasible solutions but whose solutions can represent the true minimum number of required sensors. Preliminary results herein showcase the flaws inherent in traditional methods while also establishing the basis for the application of set-based methods. The proposed methods are first developed and tested in simulation and are also developed to the point of being able to be implemented for the optimal inspection of real parts in real manufacturing environments. This includes experimental validation of the methods as well as testing in real industrial situations to demonstrate their ability to generate real, usable camera deployments. These results show agreement between theoretical simulation results and real images, and demonstrate the ability of the method to be generically applied across a wide array of different geometries and environments.