Host: Japan Society for Fuzzy Theory and Intelligent Informatics (SOFT)
In recent years, it demonstrate that concrete structure occur (happen) to decline quality by enlarged load, environment - temperature, humidity, carbon dioxide, acid rain and external pressure - materials, quality of work and it is recognized that maintenance and care are important. During maintenance and care, it is important to estimate and judge health of concrete structure. One of estimation and judgement of health, visual inspection of an expert is existent. Connecting visual information on the surface with state (situation) of declining quality inside of concrete is knowledge and experience of expert. This research focused TAM (Topographic Attentive Mapping) network as a method deal with visual information of expert. TAM network is neural network which models visual system processing from the retina down to first visual cortex and get same result which is human visual. We extract several amount of feature which based on cracking check pattern of surface of concrete board, and by using TAM network, have to master these expert knowledge by learning expert estimation and judgement. Moreover, by adopting Pruning Methods, we try to rise correct rate of TAM network and we extract special features which are factors of expert estimation and judgement from knowledge of learn.
