The neglect of perception ability to the surrounding traffic conditions has always been the major cause of traffic accidents and the inattention to blind spots is the most important factor during driving. Existing solutions are facing the problems of using expensive equipment, wrong classification of the target object type, not suitable for nighttime, and incorrectly determining if the target object is in the blind zones. This paper aims to improve driving perception ability by developing an all-day object detection and recognition system with more accurate performance for blind zones. The proposed method uses a general-purpose camera as a single input and a two-stage deep network architecture for object detection and recognition. The proposed system is based on a two-stage cascaded network structure. At first, the style conversion process is performed to convert the daytime and nighttime images with different brightness into consistent brightness. Then the objects in the visual blind zones are detected and identified. Therefore, the accuracy of object detection can be significantly improved. Due to the diversity and complexity of Taiwan’s road conditions, the public databases cannot effectively fulfill local application model training needs. Therefore, we have built training data set from available night images. Experimental results show that the proposed method has demonstrated promising performance in all-day object detection and recognition for blind zones.

This work presented a novel vision-based augmented-reality system for applications in multiuser collaborative environments. The kernel technology of this vision-based system locates the cameras that are utilized to point and simulate the positions of multiple viewers. Camera calibration based on computer vision is employed during the camera's locating process. The applications in multiuser collaborative environments allow the viewers to view various positions and in numerous directions. However, traditional calibration approaches are not suitable for these cases sufficiently. A novel calibration pattern based on pseudo-random arrays is designed for multiuser collaborative applications. The pattern has a simple and regular structure, easily extracts features, achieves robust recognition using local information, and does not limit viewer positions and directions. Experimental results indicate that the proposed system provides a effective platform for applications in multiuser collaborative environments.

In the conventional structured light approach to 3D object surface reconstruction, it is well known that finding the correspondences between the camera rays and light rays is a difficult problem. In this research we present a novel approach that does not directly use the ray correspondences to derive the surface points as the intersection of the corresponding camera rays and light rays. Instead, we use a virtual linear grid line originating at a grid junction on the object surface being considered. The length and the orientation of the virtual linear grid line in the 3D space are first calculated. Then the 3D positions of the grid junction can be determined. Detailed derivations and experimental results using the real object data are reported.

In the grid-coding technique, the accurate grid junction locations are vital to the surface computation. Noise arising from the imaging formation and processing processes will jeopardize the computation accuracy. The objective of this study is to make use of the geometric property of cross ratio to improve the accuracy of grid junction locations in grid-coded images. A method is devised to regenerate grid junctions from the original (i.e. raw) junction data of an image. Through a statistical error analysis, we show that the regenerated junction data are generally more accurate than the old ones in the image taken from a wide range of viewing angles.

A simple vision method to determine the 3-D location and shape parameters of a cone by using structured light is presented. The key concept in the proposed method is to convert the complex curved stripe pattern to an equivalent, but simpler, linear stripe pattern. The conic surface patch lying between the two visible silhouette edges is replaced by the plane defined by the two silhouette edges that cuts the cone. The appropriate geometric models are given to compute the cone axis and shape parameters representing the height of the cone and the radius of the cone base.

We are concerned with the problem of estimation of true cylinder radius, height, location, and orientation. We present mathematical models for measuring these location and shape parameters using a spatial encoding technique. A crucial step in the proposed method is to convert the estimation problem with complex curved stripe patterns to an equivalent, but simpler, estimation problem with line stripe patterns. The notions of silhouette edges and virtual plane are introduced. Various projective geometry techniques are applied to derive the cylinder location and shape parameters. The actual experiment apparatus is set up to employ the developed mathematical models to measure the cylinder geometric parameters. Description of the image-processing tasks for extracting perceived curved stripes and stripe endpoints is given. Sensitivity analysis and proper measures are taken to consider the effects of the uncertainties of stripe endpoints and silhouette edge locations on the measurement. Experiments have confirmed that our measurement method yields quite good results for different cylinders under various measurement environment conditions.>