Detecting acceleration for gait and crime scene analysis

Abstract

Identifying criminals from CCTV footage is often a difficult task for crime investigations. The quality of CCTV is often low and criminals can cover their face and wear gloves (to withhold fingerprints) when committing a crime. Gait is the optimal choice in this circumstance since people can be recognised by their walking style, even at a distance with low resolution imagery. The location of the frame when the heel strikes the floor is essential for some gait analyses. We propose a new method to detect heel strikes: by radial acceleration which can also generalise to crime analysis. The frame and position of the heel strikes can be estimated by the quantity and the circle centres of radial acceleration, derived from the optical flow (using DeepFlow). Experimental results show high detection rate on two different gait databases and good robustness under different kinds of noise. We analyse detection of heel strikes to show robustness then we analyse crime scenes to show generalisation capability since violent crime often involves much acceleration. As such, we provide a new basis to a baseline technique in crime scene analysis.

Analysing Acceleration for Motion Analysis

“…As in our previous work [3], we approximate the acceleration by differencing the velocity field between frames. By reference to the same starting position, the time axis is reversed when estimating the previous velocity 3 :…”

“…There has been little work as yet analysing acceleration before we determined gait events through acceleration flow in our earlier work [3]. Beyond that, [4] differentiated the velocity field without considering an analytic solution for isolating acceleration whereas [5] extended the brightness constraint to three frames and solved the problem in an extremely complex manner.…”

“…Beyond that, [4] differentiated the velocity field without considering an analytic solution for isolating acceleration whereas [5] extended the brightness constraint to three frames and solved the problem in an extremely complex manner. Our own previous work [3] did not include the following differential basis and was focused more on the analysis of human gait detecting in particular the heel strikes. Our new approach in this paper retains the elegance of the original Horn Schunck formulation with an approach that isolates only acceleration, thus allowing more detailed analysis of complex motion fields, and we provide a more general experimental analysis.…”

“…Our estimated detection of acceleration shows expected results on artificial scenes. We have used our new technique to detect heel strike for gait analysis in our previous work [3].…”

On parameterizing higher-order motion for behaviour recognition

“…There was little work analysing higher-order motion before our work on determining gait events through acceleration flow [34]. Chen et al [35] established an acceleration detection algorithm (without an implementation) based on Lucas-Kanade [36]: the brightness is constant in three consecutive frames and acceleration flow is smooth over a small patch, which turns the constrains into an over-determined equation.…”