Research of terrain recognition for off-road robot based on extreme learning theory

摘要

Feature extraction and classification algorithm is the key to classification accuracy. Terrain recognition for off-road robot need higher real-time classification algorithm, while the traditional neural network training method is difficult to meet the requirements. Extreme learning machine is used to classify the terrain pictures collected by robot in real time. Experimental results show that the accuracy of ELM terrain classification is slightly higher than the traditional neural network algorithm, but algorithm efficiency is raised more than a dozen times for the small sample size of 150, which meets the requirements for accuracy, especially for real time.