Estimating the speed of a car from video images / Schätzung der Fahrgeschwindigkeit anhand von Videos

Closed Circuit TV (CCTV) systems are widely used and cameras often record the road and vehicles driving by. This paper describes a case where techniques were used to estimate the speed of a car that was recorded by a CCTV... [ view full abstract ]

Closed Circuit TV (CCTV) systems are widely used and cameras often record the road and vehicles driving by. This paper describes a case where techniques were used to estimate the speed of a car that was recorded by a CCTV system. Photogrammetry is used to estimate the distance travelled by the car and knowledge of the timing of the video system is used to estimate the time interval. The recordings of reference images from a similar car driving by with known speed (controlled test drives) are used to validate the whole measurement system. The result is an estimation of the speed of the car including an estimation of the measurement errors. This is reported as a confidence interval for the speed of the car. 

The distance travelled is measured in a simplified 3D model based on a 3D laser scan of the environment. The point cloud from the 3D-laserscan is projected on the video image using reverse projection. With a 3D-model of the... [ view full abstract ]

The distance travelled is measured in a simplified 3D model based on a 3D laser scan of the environment. The point cloud from the 3D-laserscan is projected on the video image using reverse projection. With a 3D-model of the car two positions of the car in the 3D environment are derived. The distance between the two positions is the distance travelled by the car between the two video images. The timing of the video system is validated using recordings of a special clock. The resolution of the clock is 1/100 s and is especially designed to be visible in recordings of CCTV systems. The outcome of the timing is used to calculate the time between the two video images. A statistical analysis of the differences between the speed recorded during the controlled test drives and the speed derived from the video images is used to calculate the confidence interval.