Robust Object Detection in Severe Imaging Conditions using Co-Occurrence Background Model

Abstract

In this study, a spatial-dependent background model for detecting objects is used in severe imaging conditions. It is robust in the cases of sudden illumination fluctuation and burst motion background. More importantly, it is quite sensitive under the cases of underexposure, low-illumination, and narrow dynamic range, all of which are very common phenomenon using a surveillance camera. The background model maintains statistical models in the form of multiple pixel pairs with few parameters. Experiments using several challenging datasets (Heavy Fog, PETS-2001, AIST-INDOOR, and a real surveillance application) confirm the robust performance in various imaging conditions.

Haze-level prior approach to enhance object visibility under atmospheric degradation

“…The light reflected by the scenes' objects is attenuated before it reaches the camera. The emitted radiance from the object, obviously decreases with the increase in the distance between the camera and the object [1,16]. The first term in Equation 1 signifies that the intensities of the actual scene J(p) gets attenuated with the distance and is given by…”

“…Given a clear image J(p), the hazy images are synthesized using the atmospheric scattering model described in Equations 1 and 2. The atmospheric light is assumed to be pure white and hence set to [1,1,1] (RGB values). The synthetic images generated for different haze density levels (thick, moderate and thin) with t = 0.2, 0.5 and 0.8 are shown in Figure 1b-d respectively, for the actual scene shown in Figure 1a Overall, it mitigates the lack of resolution in the DCP transmission map compared to He et al [3], the level of halo artifacts were reduced around sharp edges.…”

“…Outdoor/aerially captured images are significantly devastating due to the suspended particles like dust, water droplets, aerosols, etc. [1].…”

Improvements and Experiments of a Compact Statistical Background Model

“…where γ is the adaptive lower limit, which can be determined by a computable ratio of the signal variance σ 2 pt and the random noise variance σ 2 n [11]. The location of supporting pixel depends on the character of its target pixel.…”

“…This paper reports the modification and its experimental performance of a single-Gaussian-based model: Cooccurrence Probability-based Pixel Pairs (CP3) [10,11]. CP3 is originally proposed for off-line object detection under sudden illumination changes, and it is proved to be robust in sudden illumination changes, weak illumination, and regular dynamic background.…”