Noriyuki Ozaki

Drowsiness Detection Using Facial Expression Features

This paper presents the method of detecting driver’s drowsiness level from the facial expression. The motivation for this research is to realize the novel safety system which can detect the driver’s slight drowsiness and keep the driver awake while driving. The brain wave is commonly used as the drowsiness index. However, it is not suitable for the in-vehicle system since it is measured with sensors worn over the head. We precisely investigated the relationship between the change of brain wave and other drowsiness indices that can be measured without any contact; PERCLOS, heart rate, lane deviation, and facial expression. We found that the facial expression index had the highest linear correlation with the brain wave. Therefore, we selected the facial expression as the drowsiness-detection index and automated the drowsiness detection from the facial expression. Three problems need to be solved for automation; (1) how to de ne the features of drowsy expression, (2) how to capture the features from the driver’s video-recorded facial image, and (3) how to estimate the driver’s drowsiness index from the features. First, we found that frontalis muscle, zygomaticus major muscle, and masseter muscle activated with increase of drowsiness in more than 75 percents of participants. According to the result, we determined the coordinates data of points on eyebrows, eyelids, and mouth as the features of drowsiness expression. Second, we calculated the 3D coordinates data of the features by image processing with Active Appearance Model (AAM). Third, we applied k-Nearest-Neighbor method to classify the driver’s drowsiness level. Eleven participants’ data of the features and the drowsiness level estimated by trained observers were used as the training data. We achieved the classi cation of the drivers’ drowsiness in a driving simulator into 6 levels. The average Root Mean Square Errors (RMSE) among 12 participants was less than 1.0 level.

A novel technique for raindrop detection on a car windshield using geometric-photometric model

In this paper, we describe a novel technique for detecting raindrops using in-vehicle camera images. The appearance of raindrops on a car windshield can depend on their background, so it is often difficult to detect them using conventional template matching methods, which are based on image features. Initially, we extract potential raindrop regions from images, before generating a rendered background image using a physical raindrop model based on the refraction of light rays. This rendered image is then used to identify true raindrops based on their similarity to the true background image. We propose a new model that approximates a raindrop shape as a spheroid section. This method can represent different raindrop shapes more adaptively and flexibly than conventional models, which approximate raindrops as a section of a sphere. We also extend the Maximally Stable External Regions algorithm to extract candidate raindrops and we identify three measures of image similarity using a Support Vector Machine algorithm. We conducted experiments that confirmed the effectiveness of the proposed technique.

An imager using 2-D single-photon avalanche diode array in 0.18-μm CMOS for automotive LIDAR application

A feasibility imager chip of a 32× 4-pixel array was developed in a 0.18-μm CMOS process for a small size automotive laser imaging detection and ranging. Each pixel consists of 8 single-photon avalanche diodes as a world-first 2-D pixel array with digital output macro pixel architecture which enables laser signal sensing under sunlight noise. Distance measurement results show less than 2.1% nonlinearity and 0.11-m standard deviation up to 20-m distance with 10%-refIective target under the ambient light of 75 klux.