Image analysis of insulation mineral fibres

We present two methods for measuring the diameter and length of man-made vi treous fibres based on the automated image analysis of scanning electron mi croscopy images. The fibres we want to measure are used in materials such a s glass wool, which in turn are used for thermal and acoustic insulation. T he measurement of the diameters and lengths of these fibres is used by the glass wool industry for quality control purposes. To obtain reliable qualit y estimators, the measurement of several hundred images is necessary. These measurements are usually obtained manually by operators. Manual measu rements, although reliable when performed by skilled operators, are slow du e to the need for the operators to rest often to retain their ability to sp ot faint fibres on noisy backgrounds, Moreover, the task of measuring thous ands of fibres every day even with the help of semi-automated image analysi s systems, is dull and repetitive. The need for an automated procedure whic h could replace manual measurements is quite real. For each of the two methods that we propose to accomplish this task, we pre sent the sample preparation, the microscope setting and the image analysis algorithms used for the segmentation of the fibres and for their measuremen t. We also show how a statistical analysis of the results can alleviate mos t measurement biases, and how we can estimate the true distribution of fibr e lengths by diameter class by measuring only the lengths of the fibres vis ible in the field of view.