Particle Identification By Image Processing

The paper reviews techniques for obtaining and processing images of particles and objects (irradiation, response, surveying, storage and analysis). It presents the methodologies thanks to which images can be classified and recognised (detection of domains, boundaries, shapes and textures) with a view to detecting correlations between the information provided by the images and the physical and chemical properties of the examined particles. It discusses procedures for recognizing the boundaries of images of particles, for analysing their properties (geometrical, Fourier series, fractals, etc.) and for recognizing the structures and textures of multi-component particles. The instruments used to acquire and store the images and the hardware needed to process the digitized information are described and analysed. The software needed to analyse the data (loading, color-level correction, enhancement, filtering, thresholding and labelling) are presented. Procedures for classifying vector structures that may be used to characterize images of individual particles (pattern vectors) or classes of particles (feature vectors) are discussed and those which are used to recognize particles are analysed. Applications of the above methodologies and several case studies concerning mineral grains (free and associated), polished sections of minerals, macerals, inorganic and organic materials are described.