Neural networks as applied to vision systems in recognizing 3-D objects in different orientations

The thesis is about Neural Networks as applied to Vision Systems in recognizing three dimensional objects in different orientations. It aims to capture an object, digitize and process its image into a form acceptable as input to the neural software for training and recognizing purposes. The group considers simple objects that have well defined edges. Different orientations of a particular object will be presented during the training phase of the neural network. Those orientations presented during the training phase together formed the training set. After this, the network will be presented with the same object but in a different orientation. The network must be able to recognize it as the same object even when the said object is presented in a way that is not included in the training set. The mapping network to be used for the said application is Backpropagation Network (BPN). For the hardware needed to support the neural software, the vision system, the group based it from the previous thesis on 3-D Visual Ranging and Navigating Vehicle. This thesis which has a lot of theories and concepts would serve as a basis for further application