This paper examines the key issues in constructing algorithms for processing target information in fibre-optic trunk pipeline security systems. Deep learning methods are applied in order to implement the basic functions of the environmental recognition system in the pipeline area, in particular: detecting and classifying input signals, extracting signal-event tracks, and identifying events and their sources. This study also presents a plan for walkthrough development of algorithms for processing information based on deep-learning methods, from task assignment to reference implementation and prototype testing. The basic stages of the plan are illustrated by studies which were carried out during the development of algorithms for primary signal classification in the leak detection and activity monitoring system on behalf of AO Omega, Moscow.