Shape recognition using image processing

Abstract

A Shape Recognition through Vision is a program that use for supporting industrial needs. A tracker will try to get an accurate shape for being understood by the machine. However, there are several ways to recognize the shape model prepared. The programming area specifically starts from image acquisition until image processing. The system database actually depends on the Singular Value Decomposition (SVD) where the entire pixel in the image is identified with a certain value. Therefore, this report contains a whole explanation on the program used to produce the input for the database and also how to identify the shape due to the image processing. This program will require a MATLAB program and also Microsoft excel to display the database that later on will be put inside the neural network program for output. However, the systems actually need a good database. A camera collected all the images from indoor by using the shape model during daytime. The outcome is not as expected where it is much better if only the sunlight, reflection and the unwanted shadow in the image can be removed. Real time image will not automatically remove the entire obstacle in the image. Notice that the database in Methodology chapter is all the pixel value. This is also including the obstacle that mentioned which cannot be remove even with filtering and enhancing all using Feature Extraction. Research started on all the system transform and test if the result will be the one that needed. Two demonstrations of the architecture’s implementation in MATLAB are provided - the position of an object is tracked by using the object’s shape properties and lighting condition, weather and reflection problem is solved.