Person Detection

Person detection is a core competency of FusionSystems. Our detection software stands out for its high detection performance combined with low false alarm rates. The consistent use of tracking methods provides the necessary stability of detection.

Your needs

Do you need information on whether people are present in a scene, want to know where they are and where they are moving, or want to detect posture and gestures? Then we offer you the right solution.

We work with various camera systems: colour cameras, stereo cameras for three-dimensional capture, and thermal cameras for the night. Where the task requires it, tactile sensors or laser sensors support the camera.

Technology

Machine learning with deep neural networks (deep learning) delivers reliable results for person detection. Intelligent detectors find people in the image. Semantic segmentation determines their outline with pixel accuracy and assigns body parts: head, arms, hands, legs, torso. This results in precise motion capture. Tracking follows each person through the image sequence and keeps detection stable.

Pose estimation and motion recognition

For many applications, it is not enough to know that a person is present. What matters is what the person is doing. Pose estimation determines the position of the body parts and joints and derives a skeleton model from them. The sequence of poses yields motion, posture, and gestures. Large gestures are captured just as well as very fine movements, without markers and without devices on the body.

Face detection

Face detection finds faces in the image and determines facial features such as eyes, nose, and mouth. In the vehicle, this is used to analyse the driver and other occupants, for example with regard to their perceived comfort. The same detection serves to protect privacy: faces in image and video data are automatically obscured or anonymised.

Fields of application

Typical applications of person detection are found in human-machine interaction, production and logistics, security, automotive, therapy, and healthcare.

Automotive

In the automotive sector, the detection of people is used outside and inside the vehicle. In the vehicle environment, people and other objects are detected and evaluated for driver assistance and automated driving. In the interior, face detection analyses the driver and occupants. Learn more under Automotive and Mobility.

Production and logistics

In production, the detection of people is important for automated guided vehicles, robot systems, and other machines. It serves human safety and opens up new forms of human-machine interaction: machines can be controlled with gestures, and assembly becomes easier. Learn more under Manufacturing and Mechanical Engineering.

Therapy and healthcare

Therapy and healthcare benefit from person detection above all through the capture of posture and movement. This enables natural interaction with patients. In rehabilitation as in prevention, posture and range of motion provide insight into the patient, and therapy outcomes can be documented. In work with people with physical or cognitive impairments, motion tracking encourages movement through play. One example is the MotionComposer, which translates movement into music; learn more under Healthcare and Human-Machine Interaction.

Security and public spaces

In public areas, person detection analyses the occupancy of squares and rooms and derives evaluations and recommendations from it. The analysis of movement sequences and the recognition of intentions are important features here. To protect privacy and anonymity, faces are obscured.

We describe the underlying methods under Vision AI and Sensor Fusion & Physical AI.