To date, many freight capillaries are limited or degraded. They are called “Voie Unique à Trafic Restreint » (VUTR). There is no security engagement there, which restrict traffic to one train on the track. To increase circulation, the freight line is separated into sections. For a train in section S, sections S-1 and S+1 must remain empty. To confirm the wholeness of the convoy in S, drivers must confirm the presence of rear red lights by checking visually. The procedure is time-consuming and does not bring as many improvements as expected. This paper presents a system to improve divided VUTR. On each side of the subsections, visual sensors are installed. They bring remote surveillance, easily understandable by a human agent. The video stream starts recording when the train arrives and stops when it leaves. The system also has an option to plan data acquisition on specific times, for example theoretical timetables. Data are sent on a central server, accessible to control centers, drivers, and station’s agents. This first use of this application is a human monitoring. In case of system failure, the drivers would go back to their original missions. The second use is an algorithmic processing of the data. Image processing and machine learning are considered. In both cases, the detection of red lights in section S allows to say “section S-1 is free and safe”. Note: no SIL is attributed to this system.