This paper presents a computer vision system for detecting lanes and vehicles on highway at night. This system recognizes the rear-lights of vehicles from the night scene image on highway, as viewed by a camera-assisted car behind. The detection of rear-lights is usually complicated by the existence of reflector plates on ground, multiple cars, and the environment lightening. The variation of the shape/size and height of rear-lights, and the wide range of possible distance from the observing vehicle also add extra complexity to the recognition. In this research, we take the approach of first detecting the reflector plates among the bright spots in the highway night scene, which indicate the lane borders. Through the brightness and area filtering, the reflector plates can be extracted. Removing the reflector spots from the bright-spot image and using the information of detected lanes, the pairing of rear-lights to find vehicles can be simplified. Experiment shows the success of this approach in detecting multiple vehicles on highway.

Being aware of nearby vehicles is important for autonomous vehicle or camera-assisted car running on highway. In this research, we focus on detecting/locating nearby leading vehicles and passing vehicles from the vehicle behind. By combining the detection of vehicle's underneath shadow and the confirmation of characteristic pattern, a reliable way of finding and locating leading vehicles on highway is derived. For passing vehicles in the neighboring lane, we apply a similar shadow detection scheme to detect the possible vehicle sideline on ground. Using the information of detected highway lane border, our scheme allocates an area of interests and checks the edge points density in that area to confirm the existence of neighboring vehicles. Our proposed schemes are efficient and can be run in real-time. Testing with dozens of vehicles shows successful results.

Most of the conventional approaches for facial feature detection use the template matching and correlation techniques. These kinds of approaches are very time-consuming and therefore impractical in a real-time systems. In this paper, we propose a useful geometrical face model and an efficient facial feature detection scheme. Based on the fact that human faces are constructed in the same geometrical configuration, the proposed scheme can accurately detect facial features, especially the eyes, even when the images have complex backgrounds. Experimental results demonstrate that the proposed scheme can efficiently detect human facial features and is deal for dealing with the problems caused by bad lighting condition, skew face orientation, and even facial expression.

An important use of sensing in robot systems is to initiate or terminate motions. The conditions for these actions are based on sensor data and thus called sensory conditions. The values of the sensor input may vary while the robot is moving or the environment is changing. Hence, the sensor data must be collected repetitively to support the updated evaluation of sensory conditions. In this paper, an efficient scheme for monitoring sensory conditions is presented. An auxiliary sensory processor is used to collect all sensor data (except the vision input) and monitor sensory conditions. Only the raw values of the sensor input will be checked to speed up the monitoring process. The threshold values of the raw data for matching sensory conditions are predetermined in the host computer and stored as a checking list in the sensory processor. The details of this scheme for different cases of sensory conditions are discussed. No floating point capability is needed for the sensory processor.