Hakan Kahraman: MOVING HUMAN DETECTION AND COUNTING WITH USING IMAGE PROCESSING

Computer softwares can be useful for various areas. Security is in these areas. People may be in danger in especially crowded places like metro, train stations, airports etc. Firms are spending millions of dollars to ensure people’s safety. Moving Human Detection and Counting by Using Image Processing Software provides detection and counting of moving people in a determined place. This software that is developed by Simulink Library of Matlab and it is also be useful for statistics of the crowd.

Introduction

Human Detection and Counting is developed by Simulink / Matlab. In this software, a determined video is used for making color based moving detection. Also a video device can used in this software to take an image. We used Simulink Library in the project because Simulink provides executable specifications from models, design with simulation, continuous test and verification, and implementation with automatic code generation.

Stages

1.Owerview

We have used our classmates in video. Various simulink blocks are used in this project for taking an image, converting rgb image to grayscale,applying filters, decreasing noises on the image, marking detected objects and statistics.

Simulink blocks and their connections in the Project

2-Definitions of Simulink Blocks

From Multimedia File reads video file that will done image processing.
To Video Display sends video data to your computer screen.
Optical Flow provides to the distribution of the apparent velocities
of objects in an image.
Filter includes three filters that are 2-D Mean, 2-D Mean and Median Filter consecutively. 

The contents of Filter block

2-D Mean finds mean value of each input matrix. Median Filter perform 2-D median filtering. The Median Filter block replaces the central value of an M-by-N neighborhood with its median value.

Relational Operator performs specified relational operation on inputs

Closing performs morphological closing on binary or intensity images.

Blob Analysis computes statistics for labeled regions.

Draw Shapes draws rectangles, lines, polygons, or circles on images.

Display shows the value of input.

3-Motion Detection and Counting

Taken video by From Multimedia File block is converted to intensity and it is displayed on the screen. Another link between video and Optical Flow provides us to detect motion. We used four 2-D Mean to decrease noise in image for best detection. In the stage of trial, we used autothreeshold block instead of 2-D means but we could not decrease noise in desired level. The mean block can also track the mean value in a sequence of inputs over a period of time. 

Also we used Median Filter and Closing for best noise reduction and object tracking. Then we used Blob Analysis to display counting of object. We defined specify number of blob area as 270 for best detection as shown in Figure 4

Then we used Draw Shapes block to mark detected objects in video file. Finally all results displayed on computer secreen as shown in Figure 5

You clearly see that there is one moving object in these screens and display block shows 1.