Kamiya Laboratory - Kyush Institute of Technology

Research Themes

Our main research topics are focus on various image and signal processing techniques which is obtained industrial and medical fields.

Medical image processing

We are developing computer-aided diagnosis systems based on image processing technologies such as CT(Computed Tomography), MR(Magnetic Resonance), CR(Computed Radiology), and microscope imaging. To develop the the CAD technologies, we propose some image registration algorithms, and detection of lesions using feature analysis or deep learning technique. Also we develop some segmentation methods for detection of multiple human organs. As a result, radiologists can easily detect abnormalities and treatment of diseases.

Real time video image processing

Analysis of human movement and posture is widely applied for behavior pattern recognition, sports training, and rehabilitation. In our laboratory, we develop a navigation system for automatedoperation of robot. Also, we develop an image analysis method for wheelchair to drive automatically using KINECT or a simple USB camera. We are also studying image analysis methods to detect positions for crane gripping from container images.

Signal processing

Useful information can obtain by analyzing 1-D or 2-D signal data. In our lab, we analyze the characteristics of an IGBT (Insulated Gate Bipolar Transistor; one of the power devices), which is obtained one-dimensional or image data in the manufacturing process. By using the system, real time monitoring to detect failure automatically is possible.

Robot vision

To compensate for the declining labor force caused by the declining birthrates and aging population, there is an increasing demand for factory automation and unmanned operation. Choosing a job in the logistics industry is no exception, and automation is required to improve efficiency. In this research, we develop an image analysis method that uses deep learning and conventional pattern recognition methods to detect and recognize objects from camera images attached to the robot and automatically extract the objects.

Deep learning

In recent years, artificial intelligence has been recognized as a key technology. In particular, the development of new artificial intelligence technology centered on information and communication technology (ICT) and robot technology (RT) was developed. Continue Artificial intelligence technology developed so far has achieved excellent recognition accuracy in the field of pattern recognition, but has many limitations. This is due to excessive reliance on big data and complex models. For this reason, this research aims to develop a new general-purpose intelligence recognition technology called “Beyond AI” rather than simply developing the next generation of artificial intelligence technology. In addition, using the developed intelligent learning model, we will develop analysis methods with demonstration experiments in autonomous vehicles, precision medicine, industrial robots in mind.