Development of distance to obstacle estimation system based on object recognition technique

A Jalal, Nurjazli Wada’ah (2021) Development of distance to obstacle estimation system based on object recognition technique. Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)

	Text (24 Pages) Development of distance to obstacle estimation system based on object recognition technique.pdf - Submitted Version Download (663kB)
	Text (Full Text) Development of distance to obstacle estimation system based on object recognition technique.pdf - Submitted Version Restricted to Repository staff only Download (2MB)

Abstract

Object recognition is a field of Computer Vision that detects instances of semantic objects in images or videos. As time goes by, the need in industry become wider as technologies expanded. This project proposed a system that canbe implement from object recognition technique, which entitled distance to obstacle estimation system based on object recognition system. The limitation exist on object recognition feature may lead to this project to be developed. Therefore, this project has a purpose on developing a distance to obstacle estimation system based on object recognition technique, which be able to recognise the surrounding objects and estimate their distance from the camera or client. The final output of this project is generating a voice output on displaying the status of the distance and give message to the user whether the distance measured is in safe distance or vice versa. The development proposed is using pre-trained SSD detection model trained on COCO DATASETS to obtain an accurate vision on the object recognition. The live image or video is feed onto theWeb camera, distance is measure by applying depth estimation algorithm and the input is process through the Laptop Based Server. Then a method of googleText-to-Speech is use to generate a voice modules from class of the object recognized as the final output. The accuracy of the system found to be that accurate, depends on some situations and measurements. The voice generation module would be no problem as it is only generate the voice according to the outputs and generated from certain library such as pyttsx3

Actions (login required)

View Item