Accurate Vehicle Speed Estimation From Monocular Camera Footage

Abstract

Abstract. A workflow is devised in this paper by which vehicle speeds are estimated semi-automatically via fixed DSLR camera. Deep learning algorithm YOLOv2 was used for vehicle detection, while Simple Online Realtime Tracking (SORT) algorithm enabled for tracking of vehicles. Perspective projection and scale factor were dealt with by remotely mapping corresponding image and real-world coordinates through a homography. The ensuing transformation of camera footage to British National Grid Coordinate System, allowed for the derivation of real-world distances on the planar road surface, and subsequent simultaneous vehicle speed estimations. As monitoring took place in a heavily urbanised environment, where vehicles frequently change speed, estimations were determined consecutively between frames. Speed estimations were validated against a reference dataset containing precise trajectories from a GNSS and IMU equipped vehicle platform. Estimations achieved an average root mean square error and mean absolute percentage error of 0.625 m/s and 20.922 % respectively. The robustness of the method was tested in a real-world context and environmental conditions.

Vision‐based vehicle speed estimation: A survey

“…In fact, computing the so‐called scale factor (m/px) to transform from pixel to real‐world coordinates is one of the key problems to be solved when dealing with distance and speed estimation using monocular systems. The most common features are the road/lane width or the length of a section/region manually measured [19, 24, 38, 47, 54, 64, 66, 67, 72–75, 79, 82, 84, 87–89, 98, 101, 107, 113, 128, 129], the size of a previously known object such as vehicles [27, 62, 78, 85, 90, 103, 105, 106, 110, 117] or license plates [60, 66, 68, 70, 80, 94, 133], and the length and frequency of the lane markings [14, 20, 22, 23, 26, 61, 65, 78, 108, 119, 121].…”

“…• High meter-to-pixel ratio (Figure 5a): obtained with low-medium camera resolution, low focal lengths, high camera height, covering a long road segment, multiple lanes and both directions. Some examples are [12,15,19,22,27,36,48,74,75,81,88,121,128,129]. • Medium meter-to-pixel ratio (Figure 5b): obtained with mediumhigh camera resolution, medium focal lengths, medium camera height, covering a medium road segment, multiple lanes and only one direction.…”

“…The location of the vanishing point allows the computation of the rotation matrix R. The translation matrix T is then obtained using knowledge of real-world dimensions of some object or region in the image. Another common approach to perform calibration of extrinsic parameters is using known geometric static features present on the road plane and then compute the homography transformation which includes rotation, translation and scale [31,38,47,66,67,70,72,75,79,82,86,87,98,99,110,113,119,121,128,129]. The dimensions, distances, lengths, etc., of these static features have to be known in advance.…”

“…• High meter-to-pixel ratio (Figure 5a): obtained with low-medium camera resolution, low focal lengths, high camera height, covering a long road segment, multiple lanes and both directions. Some examples are [12,15,19,22,27,36,48,74,75,81,88,121,128,129]. • Medium meter-to-pixel ratio (Figure 5b…”

A Velocity Estimation Technique for a Monocular Camera Using mmWave FMCW Radars

“…Similarly, some points of interest in the footage can be compared with already available maps or images to estimate displacement [21]. One study [22] involved tracking license plates from a camera with known parameters.…”