System design and implementation for image recognition and knowledge management based on neural network hardware

To solve problems appear in image processing with artificial neural network hardware, such as complex input of image data, difficulty in extracting image eigenvectors, knowledge management in chaos etc., a system for image recognition and knowledge management based on the KN1A artificial neural network module was designed and developed. In this system, images can be input and pretreated easily, and operation for extracting any area of images by calling KN1A hardware can be implemented conveniently. In the training process of neural networks, convergence properties can be observed in real time. Users can be freed from complex data processing and are allowed to focus on learning and parameter setting of neural networks to make the network performs better. It is proved that the system is reliability and accuracy with the license plate recognition applications.