This paper presents an example of project-based learning (PBL) in an undergraduate course on Image processing. The design of a simple, low-cost computer vision system for implementation on a chess-playing capable robot is discussed. The system is based on a standard CCD camera and a personal computer. This project is a good tool for learning most of the course material that would otherwise be mastered by homework problems and study before an exam. An algorithm which detects chess moves is proposed. It compares two or more frames captured before, during and after a played chess move, and finds differences between them, which are used to define a played chess move. Further image processing is required to eliminate false readings, recognize direction of chess moves, end eliminate image distortion. Many Image processing problems and solutions can be introduced to students, through the proposed algorithm. The results are encouraging - students without any previous knowledge in image processing and advanced topics, such as artificial intelligence (neural networks etc.), may attain a chess move recognition success rate greater than 95%, in controlled light environments.
The trajectory of a moving object may be extracted from video using image processing algorithms. However, the quality of the extracted information largely depends on the frame rate and exposure time of the camera, thus it is difficult to capture fast movement using slower and less expensive cameras. To this end, we propose to use an active modulated light source for object tracking, interacting with exposure times and subsampling existing frames. A prototype of a multi-functional active visual marker is presented in this paper. The system is based on the ESP-WROOM-32 microcontroller, which is configured to use various communication protocols, namely WiFi 802.11, RF 2.4 GHz GFSK and RS485. The microcontroller controls the RGB LED, which is used as light source. In addition, the system can be synchronized with the external real-time clock. The experimental results have illustrated the advantages and disadvantages of the designed active markers and pointed out the directions for future work and development.
Timely and accurate defect detection is essential in the leather industry, as the quality of raw leather directly impacts both the usability and value of finished products. This paper provides a systematic overview of state-of-the-art solutions and proposes a novel approach for automated detection of leather surface defects using deep neural networks based on the Inception-V3 architecture. Five defect categories are introduced, focusing on their impact on leather quality. In addition, two deep neural network architectures were analyzed and implemented for defect detection and classification: a single-channel model and a multi-channel model with arbitration. The evaluation was carried out using a combination of a custom-developed dataset and publicly available datasets, assessed with standard performance metrics. Moreover, an image annotation tool was developed to facilitate precise defect labeling and the creation of variable-size datasets. Both models demonstrated promising results on the custom dataset, achieving accuracy rates exceeding 93%. The suggested methodology enhances the research domain of leather inspection automation by creating an openly accessible image dataset, performing a comparative analysis of detection models and creating software tools for data preparation. These contributions lay the foundation for further research in leather defect detection and potential industrial implementation.
The magnitude-based Fourier descriptors (FD) are frequently used in shape-based image retrieval, due to their efficiency and effectiveness. Unlike the phase-preserving Fourier descriptors, the magnitude-based Fourier descriptors are inherently invariant under rotation and starting point change, but they discard all valuable information contained in the phase of the Fourier coefficients (FCs). In order to preserve the coefficients’ phase, the orientation and starting point of the shape must be determined. In this paper, we conducted a comprehensive evaluation of different state-of-the-art methods for determining nominal shape orientation, which can be used to extract phase-preserving Fourier descriptors: the point of maximal radius, the axis of least inertia (moments), the phase of the first harmonic, the cross-correlation, the Procrustes distance and the pseudomirror points. The methods were compared in terms of sensitivity to non-rigid transformations, retrieval performance, computational complexity and computational time. The experimental results give insight into the pros and cons of all analyzed methods.
This paper presents the design and development of a distributed measurement system for measuring pressure in high voltage circuit breakers (HV CB) and other switching apparatuses, during no-load operations. Instead of using traditional pressure transducers which require significant installation space, additional data acquisition cards and often demand for complex wiring, an in-house solution of pressure measurement is proposed. The system consists of miniature sensors, accompanied with a suitable amplifier, microcontroller unit and communication module, which may be distributed inside the interrupter unit in convenient locations. Due to the fact that the measurement values are transmitted digitally, measurement noise is significantly reduced while the wiring of the system is additionally simplified. The proposed measurement system is tested using two different interrupters (HV CB and a load break switch). The experimental results have demonstrated that the developed system is applicable, accurate, cost-effective, flexible and simple to use.