Smart Vision & and Sensing

Professor, Robotics Laboratory
Department of System Cybernetics, Graduate School of Advanced Science and Engineering
Hiroshima University
>> Research Contents
In order to establish high-speed robot senses that are much faster than human senses, we are conducting research and development of information systems and devices that can achieve real-time image processing at 1000 frames/s or greater. As well as integrated algorithms to accelerate sensor information processing, we are also studying new sensing methodologies based on vibration and flow dynamics; they are too fast for humans to sense.

High-Speed Image Recognition using HLAC Features

Real-time imagerecognition at 1000 fps is realized by implementing a parallel processingcircuit module to calculate higher-order local auto-correlation (HLAC) featureson a high-speed vision platform. The circuit module is compactly designed inorder to decrease the number of multiplications required in the HLACcalculation. The circuit module is integrated on a user-specific FPGA of thehigh-speed vision platform. The high-speed vision platform, on which the HLACcircuit module is hardware-implemented, can extract 25 HLAC features at 1000fps from 1024×1024 pixel images, which include 0th, 1st, and 2nd orderHLAC features.

In the experimental results, projected images switching at framerates as high as 250 fps are recognized by using HLAC features extracted at1000 fps on the high-speed vision platform. Two sets of experimental image patterns were used: (i) 10 numbers from 0 to 9 (maximum projection speedwas 500 fps) and (ii) 10 face images (maximum projection speed was 250 fps).
number recognition MPG movie(9.1M)
number recognition
face-image recognition (250 fps) MPG movie(3.0M)
face-image recognition (250 fps)