Keywords: image processing, pattern recognition, diagnostics, thermal deformation, metrology, edge detection, axle set measurements, machine tools.

Computer object detection and recognition based on digital images attracts the interest of many researchers. The ability of digital identification is widely studied in various disciplines and many different examples of such applications can be found in literature. For example, problems of three-dimensional object recognition are studied or application of genetic algorithm for recognition performance enhancement is discussed. From the metrology point of view, some papers deal with importance of image calibration allowing the retrieval of real world values. Among others, machine technology is one of the fields where image analysis seems to be very promising, especially when intelligent machines are considered. However, one can face some difficulties while trying to apply such analysis. First, a preliminary decision for the feasibility of solving a specific task is to be made. Then, one must decide which procedure to apply in the light of its computational efficiency and capability of solving the task considered.

This paper deals with three examples of image processing applications in machine technology. It should be emphasised that the discussion conducted in the paper focuses on applied or proposed concepts. Thus, each concept is mainly characterised and, then, exemplary results are depicted. Such an approach is dictated at the mid-stage of the research that has been conducted. In other words, the concepts are currently tested in order to estimate the possibility of their practical implementation, i.e. implementation in industrial conditions. The first example is related to the analysis of temperature fields of machine tools [1]. The concept developed is based on the parallel processing of the thermograph and digital image of the machine. The edge detection techniques are applied to the digital image [3]. The image of the detected edges is, then, overlapped on the thermographs. This allows the identification of the main elements of the machine tool in the thermographs. Consequently, such an operation facilitates the identification of the main heat sources of the machine considered. Also, it facilitates the correction of the emissivity factors that can depend on the surface state and coating (painting) analysed.

In the next example the problem of an image based wheel diameter measurement of the axle set of a railway vehicle is characterised. The aim of this experiment is to work out a measurement system, which enables measurement of the geometry of the wheels without disassembling the axle set. In order to determine the wheel diameter based on its digital image, the calibration of such an image must be done, first. This is usually obtained by applying the calibration grid [2]. In the next step the main measurements are performed. First, the edge of the wheel must be detected. The base for edge location consists of two concentric arcs. Within the region between arcs several radial segments are considered. The applied procedure searches for a sudden change in pixel intensity along these segments. If such a change is found, it is assumed that there is an edge within the analysed segment. Finally, the detected edges (coordinates of the edges) are used to estimate the circle or ellipse parameters with the minimum-square-error algorithm. Another approach has been considered, as well. It has been proposed to replace the calibration grid with the calibration wheel of a known diameter. The procedure is based on an image in which an ellipse reflects the original wheel due to an inherent perspective error. In the first step, the parameters of the ellipse are determined. Then, the ellipse centre and the minor and major ellipse axes are calculated. The points in which the minor and major axes intersect the ellipse and the ellipse centre are considered as a representation of the previously discussed calibration grid.

Finally, the paper presents one possible way of solving a task of human-machine communication. Here, image based recognition of specific gestures is implemented. The goal of our research is related to aiding diagnostics of machine tools. While diagnosing a machine tool various information is necessary for revealing reasons for malfunctioning. The machine tool operator must frequently access different databases or must communicate with expert systems that would be helpful in making proper decision. We assume that visual communication with databases or expert systems would be very desirable and could substantially simplify the operator activities.

In the conclusions of the paper some general aspects of the research conducted are highlighted. The problem of background noise and lighting conditions are discussed since changes in the background or lighting can make it difficult to perform the proposed processing and can substantially affect the final results. Also, a need for making the applied procedures more automatic is pointed out.

1

Gauszka E., Sokoowski A., "Machine tool temperature field analysis based on processed thermographs", 9th International Research / Expert Conference "Trends in the Development of Machinery and Associated Technology", TMT 2005, Antalaya, Turkey, 2005.

2

LabView 7.1, IMAQ Vision for LabView User Manual and Concept Manual, National Instruments.

3

MATLAB User's Guide, Matlab Inc, 2002.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £105 +P&P)