Foreground Detection with a Moving RGBD Camera

Koutlemanis, Panagiotis; Zabulis, Xenophon; Ntelidakis, Antonios; Argyros, Antonis A.

doi:10.1007/978-3-642-41914-0_22

Cited by 2 publications

(10 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In their work, Koutlemanis et al [31] start with a reference image comprised solely of the background, whereas in our method we do not record the background separately, rather both the foreground and the background are always visible, a more challenging scenario. Following [31], we estimate SIFT matching between two RGB frames

R_{i}

, and

R_{i + 1}

, where i is the frame number ranging from

0 \dots N - 1

. The motion characteristics of the foreground and the background vary considerably as shown in Fig.…”

Section: Methodsmentioning

confidence: 99%

“…After the registration, the 3D point cloud is back‐projected to the depth space

D_{i}^{'}

. Similar to [31], an adaptive thresholding classification C ( p ) is used to classify every point p as either the background (0) or foreground (1):

C false(p false) = ({, \begin{matrix} 1em4pt \\ 0, & if falsefalse| D_{i} false(p false) - {D^{'}}_{i} false(p false) falsefalse| \leq D_{i} false(p false) * 0.01 \\ 1, & if falsefalse| D_{i} false(p false) - {D^{'}}_{i} false(p false) falsefalse| > D_{i} false(p false) * 0.01 \end{matrix})

This classification is performed at pixel level in the depth space for each p . Thus, if the back‐projected depth value for each point,

D_{i}^{'} false(p false)

, of the registered point cloud deviates by more than 1% from the depth value

D_{i} false(p false)

, it is classified as a foreground.…”

Section: Methodsmentioning

confidence: 99%

“…The first rows shows the RGB image and the second row shows the corresponding depth frames, while the ground truth and sample results are shown in the third and the fourth row, respectively. The last row shows the results obtained from [31]. Only 30 frames were used for training, and 300 frames were used for testing.…”

Section: Methodsmentioning

confidence: 99%

“…After the registration, the 3D point cloud is back-projected to the depth space D i ′. Similar to [31], an adaptive thresholding classification C(p) is used to classify every point p as either the background (0) or foreground (1):…”

Section: Comparisonmentioning

confidence: 99%

“…The corresponding depth maps are shown in the second row. The ground truth and the results are shown in the third and the fourth rows, while the last row contains results obtained from[31] Fig. 11 Results on genSeq2 dataset: samples frames are shown in the first row.…”

mentioning

confidence: 99%

See 4 more Smart Citations

Foreground extraction for freely moving RGBD cameras

Junejo

Ahmed²

2018

IET Computer Vision

View full text Add to dashboard Cite

In this study, the authors propose a novel method to perform foreground extraction for freely moving RGBD cameras. Although the field of foreground extraction or background subtraction has been explored by the computer vision researcher community since a long time, the depth-based subtraction is relatively new and has not been extensively addressed as of yet. Most of the current methods make heavy use of geometric reconstruction, making the solutions quite restrictive. In this study, the authors make a novel use of RGB and depth data: from the RGB frame, they first extract corner features and then represent these features with the histogram of oriented gradients (HoG) descriptor. They then train a non-linear SVM on these HoG descriptors. During the test phase, they make use of the fact that the foreground object has a distinct depth ordering with respect to the rest of the scene. Hence, they use the positively classified features from accelerated segment test (FAST) features on the test frame to initiate a region growing algorithm to obtain an accurate foreground segmentation from the depth data alone. The authors demonstrate the proposed method on six datasets, and demonstrate encouraging quantitative and qualitative results.

show abstract