Abstract: A disparity calculation apparatus for a stereo camera implements ranging of an object that includes consecutive similar patterns. In stereo matching, if a plurality of corresponding point candidates are present in a sum of absolute differences or similar evaluation value distribution for a target point, an evaluation value map is generated by superimposing an evaluation value distribution of a target point, for which a plurality of corresponding points are determined to be present, and an evaluation value distribution of each other target point present in a peripheral area of that target point. The shape of an object is represented in real space around a target point for which a plurality of corresponding points are determined to be present. The true distance of a railing that extends in a straight line is determined by extracting a line segment with the strongest linearity in the evaluation value map.

Abstract: Provided are a pedestrian-crossing marking detecting method and a pedestrian-crossing marking detecting device, wherein the existence of pedestrian crossing markings and the positions thereof can be detected accurately from within a picked up image, even when detection of the intensity edges of painted sections is difficult. In the pedestrian-crossing mark detecting device (100), a road-surface distance of a predetermined range is calculated with respect to image data picked up from the periphery of the vehicle including the road, using camera installation information or a stereo camera's distance information, and the period of the pedestrian crossing markings is calculated on the basis of the road-surface distance of the predetermined range, and furthermore, a power of frequency is calculated using an even function and odd function of a square wave of the period as the basis function.

Abstract: A stereo camera implements ranging of an object that includes consecutive similar patterns. In stereo matching using the stereo camera, if a plurality of corresponding point candidates are present in a sum of absolute differences (SAD) or suchlike evaluation value distribution for a target point, an evaluation value map is generated by superimposing an evaluation value distribution of a target point for which a plurality of corresponding points are determined to be present and an evaluation value distribution of each other target point present in a peripheral area of that target point. By this means, the shape of an object is represented in real space around a target point for which a plurality of corresponding points are determined to be present, and it is possible to calculate the true distance of a railing or the like that extends in a straight line by extracting a line segment with the strongest linearity in the evaluation value map.

Abstract: An image segmenting unit (401) in the stereo image processing device (100) extracts M (a natural number between 2 and N, inclusive) number of segmented target images wherein a first partial area within a target image has been segmented into N (a natural number of 2 or more), and also extracts M number of segmented reference images wherein a second partial area within a reference image has been segmented into N. An image concatenating unit (402) serially concatenates M data strings, each comprising a intensity value from each segmented target image, to form a first image data string and also serially concatenates M data strings, each comprising a intensity value from each segmented reference image, to form a second image data string. A filtering unit (403) and a peak position detection unit (104) calculate the disparity between the standard images and the reference images.

Abstract: Provided is a stereo image processing apparatus wherein parallax can be calculated with high precision. A window-function shifting unit (411) sets a third window function, which is formed by shifting a second window function on the basis of the amount of deviation in sub-pixel units, onto an image cutting-out unit (412). The image cutting-out unit (412) applies the second window function to a position subjected to a cut-out, cuts out a unit partial target image from a standard image, applies the second window function or the third window function, and cuts out a unit partial reference image from a reference image. A peak-position detection unit (106) calculates the amount of deviation in sub-pixel units on the basis of the phase difference between a data string comprising brightness of the cut-out unit partial target image, and a data string comprising brightness of the cut-out unit partial reference image.

Abstract: An object detection device that can accurately identify an object candidate in captured stereo images as an object or a road surface. The object detection device (100) have a disparity map generator (120) that generates a disparity map based on the stereo images; a road surface estimator (130) that estimates a road surface based on the disparity map; an object candidate location extractor (140) that extracts an object candidate region above the road surface, based on the disparity map and the road surface; an object identifying region extractor (150) that extracts an object identifying region including a region around the object candidate region; a geometric feature extractor (160) that extracts a geometric feature of the object candidate based on the object identifying region; and an object identifying unit (170) that identifies whether the object candidate is an object or a road surface based on the geometric feature.

Abstract: Provided is a stereo distance measurement apparatus wherein a camera image itself is adjusted to correct the blur, thereby preventing the distance measurement time from being long, while improving the precision of disparity detection. In the apparatus (100), a blur adjusting unit (102) uses a filter to tailor the blur of one of two images used by a disparity measuring unit (103), to the blur of the other which is weaker. A camera image itself is adjusted, whereby the blur can be corrected. Moreover, tailoring the blur of the one image to the blur of the other image which is weaker allows an application of the filter, which is a simple processing, to cause the blurs of the images used in the disparity detection, to coincide with each other. Therefore, the precision of disparity detection can be improved, while the distance measurement can be performed at a high speed.

Abstract: Provided is a stereo image processing apparatus and a method of processing a stereo image, wherein calculation precision of disparity is improved, while maintaining a processing amount equal to the SAD method. In the stereo image processing apparatus (200), a data deletion unit (201) is installed in a stage prior to an image matching unit (102) and a filter unit (103), and forms a thinned-out target image and a thinned-out reference image, by thinning out a target image and a reference image. The filter unit (103) carries out filtering that uses an inverted phase filter, which is matching based on phase correlation.

Abstract: Provided are a pedestrian-crossing marking detecting method and a pedestrian-crossing marking detecting device, wherein the existence of pedestrian crossing markings and the positions thereof can be detected accurately from within a picked up image, even when detection of the intensity edges of painted sections is difficult. In the pedestrian-crossing mark detecting device (100), a road-surface distance of a predetermined range is calculated with respect to image data picked up from the periphery of the vehicle including the road, using camera installation information or a stereo camera's distance information, and the period of the pedestrian crossing markings is calculated on the basis of the road-surface distance of the predetermined range, and furthermore, a power of frequency is calculated using an even function and odd function of a square wave of the period as the basis function.

Abstract: A stereo image processor (1) performs image matching, in an image matching section (6), on partial images of the same object included in a base image and a comparative image respectively and detects pixel-precision partial displacement between the base image and the comparative image based on a result of the image matching. Next, in an inverted phase filter processing section (9) in a second matching section (8), pixel values from the comparative image are filtered with an inverted phase filter using, as a filter coefficient, pixel values from the base image with the order thereof reversed, given the partial images of the same object that have minimum pixel-precision partial displacement. Then, in a peak detecting section (10), a peak position where an output value of the filtering with the inverted phase filter is maximum is detected and sub-pixel-precision partial displacement between the base image and the comparative image is detected based on the peak position.

Abstract: Disclosed is a stereo camera that implements ranging of an object that includes consecutive similar patterns. In stereo matching using the stereo camera, if a plurality of corresponding point candidates are present in an SAD or suchlike evaluation value distribution for a target point, an evaluation value map is generated by superimposing an evaluation value distribution of a target point for which a plurality of corresponding points are determined to be present and an evaluation value distribution of each other target point present in a peripheral area of that target point. By this means, the shape of an object is represented in real space around a target point for which a plurality of corresponding points are determined to be present, and it is possible to calculate the true distance of a railing or the like that extends in a straight line by extracting a line segment with the strongest linearity in the evaluation value map.

Abstract: An abnormal action detector is provided for accurately detecting the position of an object together with an abnormal action using a cubic higher-order local auto-correlation feature. The abnormal action detector comprises a computer which generates inter-frame differential data from moving image data, extracts cubic higher-order local auto-correlation feature data on a pixel-by-pixel basis, adds the feature data for pixels within a predetermined range including each of pixels, calculates an index indicative of abnormality of the added feature data with respect to a subspace indicative of a normal action, determines an abnormality when the index is larger than a predetermined value, and outputs the position of a pixel at which the abnormality is determined. The computer further finds a subspace which exhibits a normal action from past feature data in accordance with a principal component analysis approach.

Abstract: An abnormal area detecting apparatus is provided for detecting the presence or absence and the position of abnormality with high accuracy using higher-order local auto-correlation feature. The abnormal area detecting apparatus comprises means for extracting feature data from image data on a pixel-by-pixel basis through higher-order local auto-correlation; means for adding the feature data extracted by the feature data extracting means for pixels within a predetermined range including each of pixels spaced apart by a predetermined distance; means for calculating an index indicative of abnormality of feature data with respect to a subspace indicative of a normal area; means for determining an abnormality based on the index; and means for outputting a pixel position at which an abnormal is determined. The apparatus may extract a plurality of higher-order local auto-correlation feature data which differ in displacement width.

Abstract: An abnormal action detecting device and method for detecting an abnormal action from a moving picture. An abnormal action detecting device (11) creates from-to-frame difference data from moving picture data inputted from a video camera (10), extracts feature data from three-dimensional data composed of frame-to-frame difference data by using a stereoscopic high-order local cross-correlation, computes the distance between a partial space based on a main component vector determined by a main component analysis technique from the past feature data and the latest feature data, and judges that an action is abnormal if the distance is greater than a predetermined value. By learning a normal action as a partial space and detecting an abnormal action as a deviation from the normal one, for example, even if several persons are present in the screen, an abnormal action of a person can be detected. The computational complexity is low and the real-time processing is possible.

Abstract: An abnormal action detector is provided for accurately detecting the position of an object together with an abnormal action using a cubic higher-order local auto-correlation feature. The abnormal action detector comprises a computer which generates inter-frame differential data from moving image data, extracts cubic higher-order local auto-correlation feature data on a pixel-by-pixel basis, adds the feature data for pixels within a predetermined range including each of pixels, calculates an index indicative of abnormality of the added feature data with respect to a subspace indicative of a normal action, determines an abnormality when the index is larger than a predetermined value, and outputs the position of a pixel at which the abnormality is determined. The computer further finds a subspace which exhibits a normal action from past feature data in accordance with a principal component analysis approach.