The first part of the thesis is concerned with the quality control of the detection modules and three analysis methods are studied and compared: spectral analysis, distribution function method and correlation form measurement; implementation of the spectral analysis technique is then described. The second part is concerned with the non controlled identification of the Z lines; an image processing approach is adopted which handles the peak line recognition problem as an edge detection problem. A contextual edge detection system with a-priori knowledge is presented. Examples of implementation and integration are given. 44 figs., 59 refs.