Publications

Automated reading of engraved or printed serial numbers on metallic components remains a challenging problem in industrial manufacturing due to low contrast, reflective surfaces, arbitrary orientations, and surface degradation. Conventional optical character recognition (OCR) systems often fail in such environments, particularly when text is embedded within geometrically structured objects and appears under uncontrolled rotations. This paper presents a robust end-to-end vision pipeline for accurately recognizing numeric series on industrial metallic surfaces. The proposed approach first employs an oriented object detection model to localize target objects and their associated text regions using oriented bounding boxes. The detected text areas are then rotation-normalized through geometry-aware cropping and square padding, enabling consistent downstream recognition. A YOLO-based character recognition model is subsequently applied, with dual-orientation inference used to resolve rotational ambiguity. To further improve reliability, a confidence-driven post-processing strategy suppresses duplicate detections, enforces valid text-length constraints, and selects the most consistent recognition results. Experiments on real-world industrial data demonstrate that the proposed system achieves high end-to-end recognition accuracy and robustness under challenging imaging conditions, outperforming baseline configurations without orientation handling or post-processing. The proposed pipeline offers a practical and deployable solution for automated traceability and quality control in manufacturing environments.

This paper investigates the applicability of the Pro-jection onto Convex Set (POCS)-based clustering algorithm to image compression tasks. The POCS-based clustering approach treats all data points in a given dataset as non-intersecting convex sets and performs POCS-based parallel projections from each cluster prototype onto corresponding member data points to minimize an objective function and update cluster prototypes. The POCS-based clustering algorithm has been proven to be able to yield promising results against other prevailing clustering approaches in terms of convergence time and clustering error on general clustering tasks. In this study, a comparison of various clustering schemes for image compression applications has been conducted. The evaluations and analyses on various standard test images verify that the POCS-based clustering algorithm can perform competitively against other conventional clustering methods in image compression problems.

Projection onto Convex Set (POCS) is a powerful signal processing tool for various convex optimization problems. For non-intersecting convex sets, the simultaneous POCS method can result in a minimum mean square error solution. This property of POCS has been applied to clustering analysis and the POCS-based clustering algorithm was proposed earlier. In the POCS-based clustering algorithm, each data point is treated as a convex set, and a parallel projection operation from every cluster prototype to its corresponding data members is carried out in order to minimize the objective function and to update the memberships and prototypes. The algorithm works competitively against conventional clustering methods in terms of execution speed and clustering error on general clustering tasks. In this paper, the performance of the POCS-based clustering algorithm on a more complex task, embedding clustering, is investigated in order to further demonstrate its potential in benefiting other high-level tasks. To this end, an off-the-shelf FaceNet model and an autoencoder network are adopted to synthesize two sets of feature embeddings from the Five Celebrity Faces and MNIST datasets, respectively, for experiments and analyses. The empirical evaluations show that the POCS-based clustering algorithm can yield favorable results when compared with other prevailing clustering schemes such as the K-Means and Fuzzy C-Means algorithms in embedding clustering problems.

Fire is considered one of the most serious threats to human lives which results in a high probability of fatalities. Those severe consequences stem from the heavy smoke emitted from a fire that mostly restricts the visibility of escaping victims and rescuing squad. In such hazardous circumstances, the use of a vision-based human detection system is able to improve the ability to save more lives. To this end, a thermal and infrared imaging fusion strategy based on multiple cameras for human detection in low-visibility scenarios caused by smoke is proposed in this paper. By processing with multiple cameras, vital information can be gathered to generate more useful features for human detection. Firstly, the cameras are calibrated using a Light Heating Chessboard. Afterward, the features extracted from the input images are merged prior to being passed through a lightweight deep neural network to perform the human detection task. The experiments conducted on an NVIDIA Jetson Nano computer demonstrated that the proposed method can process with reasonable speed and can achieve favorable performance with a mAP@0.5 of 95%.

A novel clustering technique based on the projection onto convex set (POCS) method, called POCS-based clustering algorithm, is proposed in this paper. The proposed POCS-based clustering algorithm exploits a parallel projection method of POCS to find appropriate cluster prototypes in the feature space. The algorithm considers each data point as a convex set and projects the cluster prototypes parallelly to the member data points. The projections are convexly combined to minimize the objective function for data clustering purpose. The performance of the proposed POCS-based clustering algorithm is verified through experiments on various synthetic datasets. The experimental results show that the proposed POCS-based clustering algorithm is competitive and efficient in terms of clustering error and execution speed when compared with other conventional clustering methods including Fuzzy C-Means (FCM) and K-Means clustering algorithms.