Tomonori Hashiyama

Active frame subtraction for pedestrian detection from images of moving camera

This paper presents a method for active background subtraction in sequential images taken from a moving camera on the vehicle. The active subtraction is carried out by estimating camera motion using a gyrosensor. Applying the proposed method to contour matching, we present a fast and robust system for detecting moving pedestrians.

Pedestrian detection using a CCD camera and gyrosensor on a vehicle

This paper presents a method of frame subtraction for eliminating a background of sequential images taken from moving camera. The subtraction is carried out by finding the difference among frame images using a gyrosensor. Applying the proposed method for contour matching, we present a fast and less erroneous pedestrian detection system.

Filter generation system with feature reconstruction using genetic algorithm

Real-time tracking of moving objects in natural scenes has taken a great role in computer vision. In order to track the targets, it is necessary to determine the object models, but it is hard to determine them in advance. Moreover, to realize real-time operations, each frame operation should be done in a short time. We propose an automatic filter construction system for tracking objects by filter processing. The filters are designed by the system itself in each environment, and are able to extract their color features which are not contained in other areas. The proposed system generates color filters which can derive feature segments using the clustering process. After that, it reconstructs the filter to obtain better performance by the genetic algorithm. We show how the filters are effective for object detection and tracking a target.

Extraction of features and attention areas in hierarchical neural networks

A new method of automatic extraction of features and attention areas from input data is proposed in this paper using a hierarchical neural network. Human-beings can distinguish useful information from those in ambiguous and noisy environments using attention functionality. It is said that human-beings may recognize information roughly at first, and pay attention to the detailed features to confirm what the information really means. The main difficulty in realizing the attention function in the computational model is to determine the part to which the system pays attention. To determine the attention areas automatically, the proposed model consists of three hierarchical neural networks, associative memory layer, middle layer and symbol layer. The associative memory plays a great role for finding out the features of the input pattern. The middle layer corresponds to the feature extracting layer. The units in each layer connect to the units in other layers. The connection weights between the layers are modified through a Hebbian learning rule. The connection weights between the input and the middle layers represent rough features of the input patterns, while those between the symbol and the input layers represent the attention areas. In the proposed model, these features and attention areas can be extracted automatically without the designers knowledge.