| Summary: | 碩士 === 遠東科技大學 === 機械工程研究所 === 100 === The nuts detection and classification system proposed by this paper is employed the neural network techniques for image recognition and detection. The nuts detection and classification system contains the image capturing, image pre-processing and image classification. In the image capturing stage, the nuts images are obtained by the CCD with IEEE 1394 interface. In the image pre-processing stage, one purpose of this stage is to remove the noise of the nuts images captured by CCD. After the noise removing process, the edge detection, Hough transform, image reduction, features enhanced, etc. will be used to obtain the features of the nuts images for the following training process. In the classification stage, the images obtained by the previous stage will be used to train the weighting parameters of the neural network. In the training process, the training parameters of neural network will be adjusted to achieve the convergence criteria. Finally, this study proposes a simple feature enhanced method to increase the recognition rate of more than 3%. From the results of the experiments, the average accuracy of the system we proposed is more than 98.90%. The time of recognition and classification is just about 0.14 seconds. These results show that the nuts detection and classification system proposed by this thesis is very efficient.
|
|---|
