International Core Journal of Engineering 2020-26 | Page 206

TABLE I
H ARDWARE RESOURCE COMPARISONS .
•
TABLE II
H ARDWARE PERFORMANCE COMPARISONS .
•
IV. C ONCLUSION
In this paper, we propose an FPGA-based moving object
detection system that can adapt to various interference en-
vironment. To meet the real-time requirement of a visual
surveillance system, we propose a pipeline architecture to
speed up the run-time performance. The frame rate of the
proposed system is 177.08 fps under the resolution of 640∗480
with clock rate 54.4MHz. Besides, a throughput as high
as 54.4MPixels/sec can be obtained. We also propose in
this paper a moving object segmentation algorithm that can
detect eight moving objects simultaneously. As can be seen
in the experiments, the interference caused by the change of
luminance, e.g., the reƀection of light due to the rain drop or
the existence of shadows, can be effectively removed, and the
moving objects can be successfully detected which justiſes
the usefulness of the proposed approach.
A. Hardware resource usage
In this paper, we implement a real-time moving ob-
ject detection system for multi-interference environment
based on ALTERA DE2-70 FPGA. The total LE (Logic
Element) used is 7,827, which is equivalent to 109,578
logic gates (each of LE is equivalent to 14 logic gates).
Besides, the total memory used is 130,434 bits, which is
equivalent to 521,736 logic gates (In general 1 bit mem-
ory can be constructed by 4 bit logic gates). Therefore,
the total gate count is 631.314. In Table I and Table
II, we compare the proposed approach with prior arts
that are also implemented with FPGA. The part with
∗ is the data of the proposed system. When reviewing
Fig. 2.
FPGA-based literatures, we can hardly ſnd a system with
moving object segmentation. This may be due to the
limited resource constrain, since this part require a lot of
hardware resources. In the proposed system, eight moving
objects can be detected simultaneously. Moreover, the
power consumption is about 257.27 mw, the clock rate is
54.4MHz, and a throughput as high as 54.4Mpixels/ sec
can be obtained.
B. Results under various interference environment
In this paper, we verify the performance of the proposed
system under various interference environment. The re-
sults are shown in Fig. 2. As can be seen in Fig. 2,
the moving objects can be successfully detected in an
environment with rain drops (upper part of Fig. 2), with
shadows (middle part of Fig. 2), and an environment of
lower luminance (bottom part of Fig. 2).
R EFERENCES
[1] E. Arbel and H. Hel-Or ”Shadow removal using intensity surfaces and
texture anchor points”,IEEE Trans. Pattern Anal. Mach. Intell., vol. 33,
no. 6, 2011, pp.1202 -1216.
[2] S. Nadimi and B. Bhanu, ”Physical models for moving shadow and
object detection in video,” IEEE Transactions on Pattern Analysis and
Machine Intelligence, vol. 26, Aug. 2004, pp. 1079-1087
[3] N. Martel-Brisson and A. Zaccarin, ”Learning and removing castshad-
ows through a multidistribution approach,” IEEE Transactions on Pattern
Analysis and Machine Intelligence, vol. 29, no. 7, July 2007, pp. 1133-
1146.
[4] R. Cucchiara, M. Piccardi, and A. Prati, ”Detecting moving objects,
ghosts, and shadows in video streams,” IEEE Transactions on Pattern
Analysis and Machine Intelligence, vol 25, Oct. 2003, pp. 1337-1342
[5] T. Kryjak, M. Komorkiewicz, and M. Gorgon, ”Real time moving
object detection for video surveillance system in FPGA,” in Design and
Architectures for Signal and Image Processing (DASIP), Nov. 2011, pp.
1-8
[6] R. Cucchiara, C. Grana, M. Piccardi, A. Prati, and S. Sirotti, ”Improving
shadow suppression in moving object detection with HSV color infor-
mation,” in Proc. IEEE Intelligent Transportation Systems Conference,
Aug. 2001, pp. 334- 339.
[7] E. Salvador, A. Cavallaro, and T. Ebrahimi, ”Cast shadow segmentation
using invariant color features,” Computer Vision and Image Understand-
ing, 2004, pp. 238-259
[8] Y. C. Chung, J. M. Wang, and S. W. Chen, ”Progressive Background
Image Generation, ” in Proc. of 15th IPPR Conf. on Computer Vision,
Graphics and Image Processing, 2002, pp. 858-865
Detection results with rain drop or shadow interference.
184