Vehicle Detection. One potential disadvantage of using the background subtraction technique for detecting vehicles is that because the background is not updated frequently, it does not account for rapid lighting changes in the scene (▇▇▇▇▇▇▇▇▇ et al., 2003). Such effects are often caused by the entrance of a highly reflective vehicle, such as a large white truck, into the scene. Before vehicles can be detected, these environmental illumination effects must be accounted for. In this current study, correction for environmental illumination effects was accomplished by using an automatic gain control (AGC). The AGC is a rectangular area that was placed in a part of the scene where the background was always visible (i.e., no vehicles were passing over the area). Thus, any changes in pixel intensities could be assumed to be due to environmental effects, since no physical objects had traversed the area. The average intensity change over this area from the background image could be determined and applied to the entire image to improve accuracy and avoid false vehicle detections: where: Δint Aagc ∑(bginti, j − iminti, j ) Δint = Aagf Aagc is the average intensity difference over the AGC area is the area of the AGC in number of pixels bginti, j iminti, j represents pixel intensity in the background image on the interval [0,1] represents pixel intensity in the foreground (current) image on the interval [0,1]. Vehicle detection was then performed with virtual detectors drawn by the user over the program scene. Each virtual detector consisted of a registration line, a detection line, and a longitudinal line, as illustrated in Figure 5-2.
Appears in 2 contracts
Sources: Research Report Agreement, Research Report Agreement