Seal matching method, system and equipment based on neural network and storage medium

• Main Authors: GONG JICHENG, TAN LIMIN, ZHAO ZHAO
• Format: Patent
• Language: Chinese; English
• Published: 06.07.2021
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; HANDLING RECORD CARRIERS; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention provides a seal matching method, system and device based on a neural network and a storage medium, and the method comprises the steps: carrying out the preprocessing of N seal sample patterns with seals and real labels, independently extracting a feature pattern matrix, and taking each color channel of each pixel as an independent convolution kernel; training a prediction network to obtain a feature matrix X and a weight matrix W and normalizing the feature matrix X and the weight matrix W; obtaining a prediction vector matrix according to the feature matrix and the weight matrix; and acquiring an arc cosine according to the real label to obtain an angle, adding a punishment angle m to all angle values corresponding to the real label, and keeping other angle values at original angles. According to the invention, under the condition of lack of enough available seal training samples, a machine learning mode can be applied to seal recognition, and the recognition rate of seal recognition is improve