The Proposed Technique

We

propose technique to optimizing a real-time panic detection. These

optimizations will include the reduction of the computational complexity while

maintaining a high accuracy. Our real-time technique does not require any prior

knowledge about a given video. Moreover, it will be applicable to various crowd

densities (high-, moderate- and low-density).

The proposed technique works in three steps:

1-

Different between two

consecutive frames:

The proposed technique starts divide

the video into frames. Then, two consecutive frames t and t?1 are compared pixel by

pixel, to found the absolute values of the differences between two

corresponding pixels. The absolute differential image is defined as follows:

Id (t, t+1) = |It+1 – It|

(1)

It is supposed to

be the value of the tth frame in image sequences. It+1 is

the value of the (t+1)th frame in image sequences. The goal of this

step to reduce the computation complexity and avoid the motion estimation. To achieve

that the result is the positive number that represents the dissimilarity between

pixels, if there is different that mean there is motion.

2-

Calculate the Wavelet Transforms

The result from the

previous step is a positive number used as input of Wavelet Transforms 1.

A brief description of Wavelet Transforms in the following section.

Images as smooth regions interrupted by edges, these edges are often the most interesting

parts of the data both perceptually and in terms of the information they

provide. So, wavelet transform is a powerful tool for data analysis and present

the edges. A wavelet from the name is

small waves with limited duration. The wavelets transform is set of mathematical

function used to decompose data into different component. The image will divide

into four different subbands as LL (Low frequency), HH (high frequency

diagonal), HL (low frequency horizontal) and LH (low frequency Vertical). This

breaking process can be repeated to have multi-level wavelet components like 2Level,

3Level. In our proposed technique we applied the three level 1 ,2 and 3 of wavelet

transform. Coefficients matrices cH, cV, and cD

(horizontal, vertical, and diagonal, respectively), obtained by wavelet

decomposition of the input which is different between two consecutive frames. In principle after comparative study between

the results, we found the level 3 is better than other levels because it gave

more accurate result with decrease the false alarm. The goal of this step to found

high frequency part which mean the possibility of panic event is increase.

3-

Sum of all coefficient

After calculate the Wavelet Transforms

in the different between two consecutive frames. We divide each subband into one

block and four blocks to sum the coefficient cH, cD and cV on each block for

all frames. The result of this step we will have the sum of coefficient cH1 , cH2

and cH3 for all frames in case one block and four blocks. The same process

repeated for the cV and cD coefficients. When the value of the sum the coefficient

is high compare to the all previous frames that’s mean there is panic event. The following equation is used for sum of coefficients:

S =

(2)

Where S denote to sum of coefficient,

n number of frames, and c is the coefficient. The goal of this step to detect

the panic event.