A DSM camera against driver fatigue

This pairing divides the roles. The camera looks at the face and decides for itself what has happened. The FMC650, a PROFESSIONAL series tracker, adds coordinates, speed and time to the event and sends it all to a dedicated fleet management server. The manager sees not an abstract breach of the rules but a specific point on the map, a moment in time and the evidence.

What the camera recognises. The built-in warning scenarios are drowsiness, yawning, the eyes leaving the road, smoking, talking on the phone and an unfastened seat belt. In addition, DSM recognises the driver's face: up to 990 profiles fit in its memory, so the system knows exactly who has taken the wheel, with no cards or keys.

Recognition is built on deep learning algorithms. The camera is designed to work by day and by night and can recognise a face in a mask or a hat, but the quality of recognition depends on the mounting angle, on glare and on dark glasses, so it should be tested on your own drivers. It tracks the position of the head in space, the direction of gaze and how far the eyes and mouth are open. Its conclusions about drowsiness or distraction are drawn from that data.

Fatigue. The most dangerous part is that the driver often does not notice how tired he is. The camera triggers if the eyes stay closed longer than a set time or close repeatedly within 30 seconds. Yawning is tracked separately: if the driver yawns several times in a row within a minute, a voice warning sounds. That is a signal to stop and rest before something worse happens.

The belt. If the vehicle is travelling faster than 10 km/h and the driver is not belted in, the camera warns by voice. The signal repeats until the belt is fastened.

The phone. A conversation with the phone held to the ear is one of the most frequent causes of distraction. While the driver keeps talking, the camera reminds him every 30 seconds.

Smoking. Smoking at the wheel is not usually forbidden by law, but many companies ban it in their vehicles: it is a distraction, it leaves a smell in the cab and it puts the load at risk. The camera notices a cigarette and reacts after a short interval, about 2 to 4 seconds.

What the fleet manager sees. Every event goes to the server: the type of breach, the driver, the place, the speed, a still or a clip. From these, reports are built for each driver and for the fleet as a whole: who is distracted by a phone most often, who shows repeated signs of fatigue, on which runs and at which time of day. This data can be used in an incentive scheme, in accident investigations and in driver training.

The camera gives the voice warning itself, in the cab, with no dispatcher involved. There is one condition: DSM has to be calibrated to the driver's seating position, and the thresholds and volume set so that the signal can be heard with the engine and the air conditioning running. The driver does not need to look at a screen, he only needs to hear the phrase. There is a limitation too: the signal does not stop the vehicle and does not guarantee a reaction, so repeated fatigue events within a shift should reach the dispatcher as an alert rather than only appearing in the evening report.

An example from real life. A bus driver starts his third run of the day after lunch. On a straight stretch of main road the camera detects that his eyes are closing for longer and longer and gives a voice signal twice. The driver pulls into the nearest lay-by. That evening the manager sees two fatigue events on the same stretch in the report and reworks the roster so that this driver does not have three long runs in a row.

How to launch DSM in a fleet without the drivers going to war with the camera. For the first two or three weeks it makes sense to work in observation mode: the camera warns the drivers, events accumulate on the server, but nobody is penalised. During that time it becomes clear which thresholds are too sensitive and produce false triggers, and which breaches are genuinely widespread. Then a short meeting is held with the drivers, the statistics are shown without names and the rules are explained: what will lead to a review and what will earn a bonus.

After that the work becomes routine. Once a week the manager looks at the report, picks out the few most frequent offenders and goes through specific clips with them. Experience of in-cab camera rollouts usually shows the same thing: the number of events falls noticeably in the first months, once drivers understand that the system really works and that its data is used.

Combining this with tracker data broadens the picture. The FMC650 itself records harsh acceleration, braking and cornering, speeding and engine operation. If a phone event sits next to an episode of harsh braking, that is no longer a formal breach of the rules but a dangerous situation worth reviewing first.

A word about night runs. Most fatigue episodes fall in the early morning and the second half of the night, when the body wants to sleep regardless of how much rest the person had during the day. DSM reports by time of day help you see this in your own fleet and rework rosters so that the longest stretches do not fall in those hours.

Fitting and calibration. Before launch, the firmware version of the camera and the tracker and the RS232 port configuration are checked: they determine which events are transmitted and with what attachment. After fitting, the camera is calibrated on a real driver in his usual seating position, and then a short test drive is made in which each event is triggered deliberately: phone to the ear, head turned away, belt unfastened. Acceptance counts as passed when every event tested has arrived on the platform with a photo or video, coordinates and a time.

What an event record looks like. In the report it is one row: event type, driver, date and time, speed, address or point on the map, a link to a still or a short clip. Next to it you can see the tracker events for the same seconds, harsh braking for instance.

The response procedure is best written down in advance. A single phone event — a review at the weekly meeting. Two or more fatigue events within an hour — the dispatcher rings the driver and a decision is taken about stopping. A camera taped over, or events going missing — a check the same day. The specific thresholds are chosen by each fleet after the observation period.