Detection Of Anomaly Object By Using Humanoid Robot

Mansor, Nuratiqa Natrah (2016) Detection Of Anomaly Object By Using Humanoid Robot. Project Report. UTeM, Melaka, Malaysia . (Submitted)

Text (24 Pages) Detection Of Anomaly Object By Using Humanoid Robot.pdf - Submitted Version Download (678kB)

Abstract

On the way to identify the anomaly behaviour, the main issues are object detection, position and the algorithm on how to point out the target that being digress from the normal situation. Many researches out there are about the anomalies in video surveillance in a crowd scene but there is least about the abnormal object in a frame with considering the various position and different object size. Thus, this project introduced a technique for a humanoid robot to detect and learn the representations of objects within a control environment but at the same time detecting the anomalies pattern that may happen when the object is partially placed out from the respective frame and the size limitation of object from the smallest until the biggest size. It involves the application of vision system to computes the shape, body colour and pixels values to accurately identify the location and pattern of anomalies. Altered scenario in series of experiments are when another frames are put next to the current center frame, different position of object and object size are unlikely similar in every frame. Briefly, all variables and the performances of the experiment will be observed and tabulated. The project have been showcased by using robotic platform software to test the performances of the anomalies and normal form in different situation. With four different experiments, the anomaly condition and normal condition can be proved and distinguished. The actual location of anomaly in one frame can be located using mathematical formulae too. By mean of that, the mean value for anomaly is decreasing about 50% from the normal frame and majority of the values for standard deviation are twofold from the common condition when the size of object is manipulated. The limitation for object size is starting from 0.4x0.4 cm until 50x50 cm. While, the mean values had minor difference about 0.01-0.04 from the normal one and the deviation values are fluctuated by way the effects of object position proposed into the system. The deviation value when the object is located between two frames are remarkably the lowest compared to the value when full object is totally in a frame.

Actions (login required)

View Item