Tomislav Matić

CPU, GPU and FPGA Implementations of MALD: Ceramic Tile Surface Defects Detection Algorithm

This paper addresses adjustments, implementation and performance comparison of the Moving Average with Local Difference (MALD) method for ceramic tile surface defects detection. Ceramic tile production process is completely autonomous, except the final stage where human eye is required for defects detection. Recent computational platform development and advances in machine vision provides us with several options for MALD algorithm implementation. In order to exploit the shortest execution time for ceramic tile production process, the MALD method is implemented on three different platforms: CPU, GPU and FPGA, and it is implemented on each platform in at least two ways. Implementations are done in MATLABs MEX/C++, C++, CUDA/C++, VHDL and Assembly programming languages. Execution times are measured and compared for different algorithms and their implementations on different computational platforms.

Parallel processing with CUDA in ceramic tiles classification

This paper describes the implementation of an algorithm for surface error detection on ceramic tiles in CUDA (Compute Unified Device Architecture). It compares the differences between the CPU and the GPU algorithm implementation, analyzes the features of CUDA GPU and summarizes the general programming model of CUDA. Paper presents the speed up gained in favor of the GPU algorithm implementation. Implemented algorithm used in this paper written in C is relatively simple, and for test results version for the CPU was made and the GPU version. The results show the speed up of the computation compared with the CPU that increases as the image size increases, with the maximum speed up of 4,89 times.

Technology transfer of computer vision defect detection to ceramic tiles industry

Visual inspection is carried out manually in ceramic tile industry in Croatia using specially trained and skilled workers. Currently in industry biscuit and crude tiles are not visually inspected for defects. Fatigue, illness and other subjective factors significantly influence workers percentage of found defects and classification quality. In this paper we present a prototype computer vision station (CVS) for real-time biscuit tile defects detection. CVS is a result of an FP7 project. Prototype is mounted on a production conveyor line before the kiln. MFC (Microsoft Foundation Class) based GUI application is created and all developed algorithms are implemented in C++ language using OpenCV and Nvidia CUDA libraries. System hardware is based on core i7 CPU and Nvidia GTX960 GPU. Preliminary results show maximum execution time below 900 ms and defect detection efficiency of 98%.

Real-time Biscuit Tile Image Segmentation Method Based on Edge Detection

In this paper we propose a novel real-time Biscuit Tile Segmentation (BTS) method for images from ceramic tile production line. BTS method is based on signal change detection and contour tracing with a main goal of separating tile pixels from background in images captured on the production line. Usually, human operators are visually inspecting and classifying produced ceramic tiles. Computer vision and image processing techniques can automate visual inspection process if they fulfill real-time requirements. Important step in this process is a real-time tile pixels segmentation. BTS method is implemented for parallel execution on a GPU device to satisfy the real-time constraints of tile production line. BTS method outperforms 2D threshold-based methods, 1D edge detection methods and contour-based methods. Proposed BTS method is in use in the biscuit tile production line.

Semi-automatic prototype system for bacterial colony counting

Bacterial colony forming unit (CFU) counting is a tedious task mostly done by humans. Procedure is error prone, time-consuming and laborious. In this paper we present a semi-automatic prototype system for CFU counting. The developed prototype consist of a hardware (area scan camera with LED light source) and C#-based desktop application. The application enables manual, semi-automatic and automatic CFU counting. Automatic CFU counting is based on Hough transform for circles. Obtained results can be user-corrected for better accuracy. In the experimental analysis, the developed application is evaluated on the synthetic CFU images. The results include time and counting performance measurements compared with the manual count. The results show that semi-automatic counting procedure can save on average 45% of counting time compared to manual count with the same counting accuracy. The automatic CFU counting on average has precision of 97% and recall of 82%.