Abstract: The materials such as wood, cast and steel products, because their density are more homogeneous, so their image profiles which are inspected by ICT device must be binary-value images approximately. Based on this prophetic knowledge, an optimization strategy of maximum image-entropy and neural networks is adopted in image reconstruction algorithms, the output functions of neural cells is modified too. Moreover, the image illumination is fixed before and after image reconstruction. The results of experimentation show that a high resolution defects image of homogeneous material can be reconstructed under only 6 projection angles of ICT device. So, there will be a good approach to improve the work efficiency and reduce the cost of ICT devices.

Key words: Defects exploration, Homogeneous material, Industry CT, Maximum image-entropy, Neural network reconstruction algorithm