Rotated object recognition-based on Hu moment invariants using artificial neural system
Amitabh Wahi, C. Palamsamy, S. Sundaramurthy · 2012
This paper presents eight class rotated objects recognition by employment of supervised feedforward neural network (FFNN). The process consists of two parts. First, segmentation of the binary edge object from the colored image is carried out. The binary edge image is rotated at the center of the image from 0 degree to 360 degree by every 5 degree rotation. The invariance features are extracted by Hu moments from each image. The same process is repeated for rest of the images. In the second phase, the seventy five percentages of randomly selected data are presented to train the two hidden layers neural classifier respectively. The performance of back error propagation trained neural classifier is evaluated on twenty five percentages randomly data set. It is found that 94% of classification accuracy is obtained on test set by FFNN.