Chikahito Nakajima

Full-body person recognition system

We describe a system that learns from examples to recognize persons in images taken indoors. Images of full-body persons are represented by color-based and shape-based features. Recognition is carried out through combinations of Support Vector Machine (SVM) classifiers. Different types of multi-class strategies based on SVMs are explored and compared to k-Nearest Neighbors classifiers. The experimental results show high recognition rates and indicate the strength of SVM-based classifiers to improve both generalization and run-time performance. The system works in real-time.

People recognition and pose estimation in image sequences

Presents a system which learns from examples to automatically recognize people and estimate their poses in image sequences with the potential application to daily surveillance in indoor environments. The person in the image is represented by a set of features based on color and shape information. Recognition is carried out through a hierarchy of biclass SVM classifiers that are separately trained to recognize people and estimate their poses. The system shows a very high accuracy in people recognition and about 85% level of performance in pose estimation, outperforming in both cases k-nearest neighbors classifiers. The system works in real time.

Object recognition and detection by a combination of support vector machine and rotation invariant phase only correlation

This paper proposes an object recognition and detection method by a combination of support vector machine classifier (SVM) and rotation invariant phase-only correlation (RIPOC). SVM is a learning technique that is well founded in statistical learning theory. RIPOC is a position and rotation invariant pattern matching technique. We combined these two techniques to develop an augmented reality system. This system can recognize and detect objects from image sequences without special image marks or sensors and show information about the objects through a head-mounted display. Performance is real time.

Extraction of salient apexes from an image by using the function at the primary visual cortex

The mechanism of human vision has become clearer as a result of physiological study. So this paper proposes a method, which is based on the structure and function of primary visual cortex in the human brain, to extract salient apexes from an image. It extracts potential apexes by using the local gradient of brightness. Also it selects salient apexes from the potential apexes by using the functions of the cells which show the inhibited response to long lines and the activated response to different directions of lines. This method can extract the features from not only images but also drawings, such as drawings of characters, electric circuits, facilities, etc.

Maintenance Training of Electric Power Facilities Using Object Recognition by SVM

We are developing a support system for maintenance training of electric power facilities using augmented reality. To develop in the system, we evaluated the use of Support VectorMachine Classifier (SVM) for object recognition. This paper presents our experimental results of object recognition by combinations of SVMs. The recognition results of over 10,000 images show very high performance rates. The support system that uses the combinations of SVMs works in real time without special marks or sensors.

Feature detection based on directional co-occurrence histograms

This paper proposes a novel feature-detection method based on the co-occurrence histogram for an image. The method uses eight co-occurrence histograms to emphasize features in the image. The image features are treated as occurrence frequencies of two-pixel combinations in the image. When the proposed method is applied to a drawn image like a rectangle, the corner points are emphasized in the image. When the proposed method is applied to a gamma irradiation image, the noises caused by the irradiation are emphasized in the image. The experimental results indicate that the method can emphasize rare features in the images. The proposed method can be applied as a pre-process of feature detection in an image processing system.

Autodetection of barnacle larvae at power plants

Barnacles are one of the major fouling organisms at electric power plants. They cling to the inside of pipes and the surface of steam condensers of the power plants, thus reducing the stream of seawater into the power plants. Furthermore, they erode the pipes and the steam condensers. A continuous auto-detection system of barnacle larvae is necessary to decide the best time for using antifouling techniques. This paper describes a continuous auto-detection system for barnacle larvae and provides experimental results for larvae detection at the sluice gate of a power plant.