بناء مقياس مصور للسلوك العدواني لتلاميذ المرحلة الابتدائية بعمر 8 - 7 سنوات للحد من استخدام الالعاب الصغيرة == Design And Implementation of Image Tracker Algorithm Based On Matlab
Author name:
مهند عبد كيطان
Supervisor name:
غازي صالح محمود
General topic:
Sports
Specific topic:
Physical Education
Degree:
Master
University:
Mustansiriyah University - Faculty Of Basic Education - Department Of Physical Education
Language:
Arabic
University location:
Baghdad
First pages:
17T1528 - p.pdf
Abstract:
Trithm is designed to serve two consecutive surveillance stations, in the first station; a pre - recorded video is used to select a vehicle of interest, interac - tively by an operator. The morphological properties of selected vehicle are to be transmitted to the following station and used to recognize the selected vehicle.When the objects pass in front of the second camera viewing area, the algorithm compares between each object?s measured property element and the corresponding received one from previous station. When all proper - ty elements agree to within an absolute tolerance, the object is declared a matching one by bounding it with a red box and showing its path in dotted red line. This treatment continues until the object exits the second camera viewing area (being outside the region of interest, ROI). Tracking process in the second station is performed in an on - line manner, as there is no pre - tracking data.A fixed camera has been considered for motion detection. By sub - tracting the background model from current image, the difference image of variant pixels is generated. Otsu's thresholding method has been used to find a suitable threshold to filter out the camera noise and to determine the sensitivity of motion detection. Morphological operations have been used to improve the segmentation results.Compensation for digital camera's white - balance algorithm has been used (subtracting image mean value) prior to generating difference image. This treatment has satisfyingly enhanced the segmentation operation. Moreover, using Euler number as a feature of a connected area, and adjust - ing the morphological operation sequence to be (close, thicken) combina - tion, has greatly enhanced recognition process; since thicken operation pre - serves Euler number, and hence keeping an important object feature.Adding salt and pepper noise to video frames has shown rather high algorithm sensitivity against noise, since it (the noise) directly affects con - nected area features, such as area and Euler number; 5 percent of noise has reduced hit ratio to about zero percent.The time performance three detection algorithms, namely temporal differencing, fixed background, and Gaussian mixture models, has been tested for a video of (120 x 160) pixels frame - size. For computational cheapness, the fixed background algorithm (which is based on taking the first captured empty video frame as a background) has been adopted in this thesis.
