Image Processing Reference
In-Depth Information
4.4 Area Motion Detector
We can modify the Over a Line Motion Detector algorithm and instead of specifying one line,
we can specify multiple lines. We can specify several detached areas or one area defined by a
polygon. This algorithm can be used easily to define an area of interest or exclude the defined
area; this can be done by simply reversing the order of the points that define the area. We will
need to compute N number of dot products where N is the number of lines.
4.5 Fire Detection
As a next step, we wanted to detect not only motion but also some type of an emergency
event such as fire. If we classify the pixels that contain fire, then we can highlight these pixels
(change their color, enhance their brightness, etc.) and also notify an operator. This is a dif-
icult problem to solve as it is not very trivial to classify and recognize pixels that contain
fire, flame, or smoke. We used Refs. [ 33 , 34 ] to figure out how to detect fire in a video stream.
Then we implemented the algorithm in Ref. [ 33 ] which produced relatively accurate results
as shown in Figure 4 . The angle of the camera seemed to influence the fire detection, but if
the fire was close to the camera, this algorithm seemed to produce accurate results. We should
note here that we did not incorporate the "Compute Area" part of the algorithm presented in
[ 33 ] and that could be the reason of not having our implementation producing very accurate
results.
FIGURE 4 Fire detection in a video stream.
5 Conclusion
In this article, we presented several algorithms that are easily implemented even by under-
graduate students. These algorithms have many applications in Security and more speciic-
ally in security surveillance. Using CUDA, these algorithms can be executed in parallel for in-
creased performance. This can be done because using CUDA we can launch one thread per
pixel. So, a high-resolution image or video feed can be processed in almost the same amount
of time as a low resolution image or video stream. When our students implemented these al-
gorithms as programming assignments we found that the biggest delay in the application was
the network feeding the camera video stream to the application. In this paper we present the
extended results of our paper initially published in Ref. [ 35 ] .
References
[1] Olsina L, Dieser A, Covella G. Metrics and indicators as key organizational assets
for ICT security assessment. In: Akhgar B, Arabnia HR, eds. Emerging trends in com-
puter science & applied computing. Elsevier Inc.; 2013:978-0-12-411474-6. Emerging trends
in ICT security.
 
 
Search WWH ::




Custom Search