Character recognition for automotive electrical box components based on Machine vision

摘要

According to the features of the electric box components, the characters on the components are identified from the general optical character recognition process. The problem of multi classification in the recognition process is studied. First, the paper introduces the machine vision components of automotive electrical box system based on automatic detection; then, according to the multi classification support vector machine learning, we propose a support vector machine and active learning and give the idea of combining active learning model, the inquiry mechanism is established according to the fuzzy sample in the learning process; Finally, a multi-classifier construction algorithm is given based on the interrogation strategy. The algorithm is applied to character recognition of new energy electrical box components by software programming, and the effectiveness of the algorithm is verified, the robustness of the algorithm is enhanced, and the recognition rate is improved.