A computer vision based new method for the measurement of the length of pedestrian crossings using a single camera is described. The main objective of this research is to develop a travel aid for the blind people. In a crossing, the usual black road surface is painted with constant width periodic white bands. In Japan, this width is 45 cm. The crossing region as well as its length is detected using this concept. At first, the crossing direction is determined from the power spectrum using fast Fourier transform. The periodic white and black bands are detected using integration along the crossing direction and then differentiation of the integral data perpendicular to crossing. This detection may be erroneous due to adverse effects of the neighboring region of crossing, as the intensity of the whole image is used for bands detection. To remove the neighboring effects, the crossing region is extracted. Then the crossing bands are detected from the image intensity using the crossing region only. Experiment is performed using 32 real road scenes with pedestrian crossing. The rms error is found 2.28 m. The technique determines the crossing length with good accuracy for crossings marked clearly with white paintings as well as fine image resolution.
Stereo correspondence is a common tool in computer or robot vision, with numerous applications, such as determination of three-dimensional depth information of objects for virtual reality, autonomous vehicle and robot navigation, using a pair of left and right images from a stereo camera system. Computation time is an important factor in estimating dense disparity for the above applications. For of a pixel in the left image, its correspondence has to be searched in the right image based on epipolar line and maximum disparity search range. The intensity of a pixel alone in the left image does not have sufficient discriminatory power to determine its correspondence uniquely from the right image, thus other pixels in its neighborhood comprising a window is used for accurate estimation. In window-based approaches, this correspondence or disparity is conventionally determined based on matching windows of pixels by using sum of square differences, sum of absolute differences, or normalized correlation techniques. With a view to reduce the computation time, we propose a fast algorithm where it is not necessary to compute the window costs for all candidate pixels in the right image within the search range. To determine the correspondence of a pixel in the left image we just compute the window costs for candidate pixels in the right image whose intensities are different within a certain value to the intensity of the pixel in the left image. We applied our proposal to standard stereo images and found that we can easily reduce the computation time of about 30% with almost no degradation of accuracy.
Debris flow causes lot of damages all over the world. The surface velocity of such a natural random flow is one of the important physical parameters that are considered in drawing a hazard map or constructing a sediment control dam. Gradient-based method along with a multiscale smoothing operation finds application in the two-dimensional velocity estimation of a debris flow from its video images. This paper investigates the performance of another optical flow determination technique - the cross-correlation method in the above application and compares the results with those obtained by the gradient-based method. A computer simulation with synthetic random moving images shows that the accuracy of the cross correlation method is higher than that of the gradient-based method.