Abstract: A method and a system for localizing at least one circular object in a video frame captured by an image capture unit are disclosed. The method includes the steps of: A. performing edge detection to obtain a plurality of image edge points in the video frame; B. extracting a contour set of image contours by linking neighboring edge points; C. generating a plurality of circle samples from the extracted contour set by a structural sampling process; and D. localizing the at least one circular object from the generated circle samples by calculating a respective fitness score for each of the circle samples to measure the supporting evidence of the image edge points on a neighborhood of each of the circle samples, and by choosing the circle sample(s) whose fitness score(s) is/are greater than a threshold value as the circular object(s) found in the video frame.

Abstract: A method and a system for localizing at least one circular object in a video frame captured by an image capture unit are disclosed. The method includes the steps of: A. performing edge detection to obtain a plurality of image edge points in the video frame; B. extracting a contour set of image contours by linking neighboring edge points; C. generating a plurality of circle samples from the extracted contour set by a structural sampling process; and D. localizing the at least one circular object from the generated circle samples by calculating a respective fitness score for each of the circle samples to measure the supporting evidence of the image edge points on a neighborhood of each of the circle samples, and by choosing the circle sample(s) whose fitness score(s) is/are greater than a threshold value as the circular object(s) found in the video frame.

Abstract: An efficient 3D object localization method using multiple cameras is provided. The proposed method comprises a three-dimensional object localization process that firstly generates a plurality of two-dimensional line samples originated from a pre-calibrated vanishing point in each camera view for representing foreground video objects, secondly constructs a plurality of three-dimensional line samples from the two-dimensional line samples in all the multiple camera views, and thirdly determines three-dimensional object locations by clustering the three-dimensional line samples into object groups.

Abstract: A video object detection system based on region transition includes a video acquiring unit, a user interface unit and a control module. The video acquiring unit is utilized for acquiring a video frame. The user interface unit is configured to provide a user to define at least one detection region with a set of image pixels on the acquired video frame and to define at least one region transition rule for identifying video objects of interest. Each detection region is represented with a set of image pixels. The control module is utilized for detecting position of a target video object and determining whether a moving trajectory of the target video object matches the at least one region transition rule.

Abstract: A background model learning system for lighting change adaptation for video surveillance is provided. The system includes a background model estimation unit that estimates a background model for a scene of interest; a foreground map construction unit that constructs a reference foreground map of the time instance; and a lighting change processing unit that revises the reference foreground map by reducing false foreground regions resulting from lighting changes. The revised foreground map is then sent back to both the background model estimation unit and the lighting change processing unit as feedbacks for model learning rate tuning in background model estimation and map integration in lighting change processing, respectively, for the next time instance.

Abstract: In the present invention, we identify that such a tradeoff between robustness to background changes and sensitivity to foreground abnormalities can be easily controlled by a new computational scheme of two-type learning rate control for the Gaussian mixture modeling (GMM). Based on the proposed rate control scheme, a new video surveillance system that applies feedbacks of pixel properties computed in object-level analysis to the learning rate controls of the GMM in pixel-level background modeling is developed. Such a system gives better regularization of background adaptation and is efficient in resolving the tradeoff for many surveillance applications.

Abstract: An efficient 3D object localization method using multiple cameras is provided. The proposed method comprises a three-dimensional object localization process that firstly generates a plurality of two-dimensional line samples originated from a pre-calibrated vanishing point in each camera view for representing foreground video objects, secondly constructs a plurality of three-dimensional line samples from the two-dimensional line samples in all the multiple camera views, and thirdly determines three-dimensional object to locations by clustering the three-dimensional line samples into object groups.

Abstract: To construct a video surveillance system, the Gaussian mixture modeling (GMM) is a popular choice among various background modeling approaches, for its capability of adaptation to background variations. However, the GMM often suffers from a tradeoff between robustness to background changes and sensitivity to foreground abnormalities, and is inefficient in managing the tradeoff for diverse surveillance scenarios. In the present invention, we identify that such a tradeoff can be easily controlled by a new computational scheme of two-type learning rate control for the GMM. Based on the proposed rate control scheme, a new video surveillance system that applies feedbacks of pixel properties computed in object-level analysis to the learning rate controls of the GMM in pixel-level background modeling is developed. Such a system gives better regularization of background adaptation and is efficient in resolving the tradeoff for many surveillance applications.

Abstract: Surveillance systems often encounter great challenges from lighting variations, especially for those inspecting outdoor environments. To construct a surveillance system robust to various background scene changes, including lighting variations, a strategy of background model learning is widely adopted. Based on this strategy, many approaches have been proposed in decades to represent background scenes by statistical models and to adapt background changes over time into the models. However, the focus of most background model learning research is put on adaptation of scene vibrations in to background, as well as of gradual lighting variations. For the background model adaptation to drastic lighting changes, many background model learning approaches are often inefficient. As a result, false alarms in foreground detection are issued under such quick lighting changes. To suppress this kind of false alarms, a new system design of background model learning is proposed.

Abstract: A method for adjusting photographing settings of a digital camera through motion detection, for determine to adjust the photographing settings of the digital camera according to vibrations of the digital camera or shaking of a photographed object. The adjusting method includes the following steps. At least two continuous pre-photographed images are captured. A central area and an edge area are set at same positions in the pre-photographed images respectively. Feature points are found in the pre-photographed images. A local motion vector and a camera motion vector are respectively determined according to a relation of the feature points on the central area and the edge area. A subject motion vector is generated according to a difference between the local motion vector and the camera motion vector. Adjustment of the photographing settings of the digital camera is determined according to a maximum value in the subject motion vector and camera motion vector.

Abstract: An image processing device includes an image acquisition module, a memory module, and an image signal processing module, for performing an image enlargement and enhancement. The image acquisition module sequentially reads in an image block, including a unit pixel matrix and an exterior pixel matrix, wherein each pixel matrix includes a plurality of pixels and each pixel is associated with a parameter. The memory module stores a plurality of predefined edge patterns. The image signal processing module compares a loaded image block with predefined edge patterns, and determines if it is an edge block. Then, the image signal processing module further classifies its pixels into two groups, and calculates a continuous separating boundary between the two groups. Finally, the image signal processing module enlarges an edge block by placing new pixels inside its unit pixel matrix, wherein the new pixel parameters are extrapolated from the two classified pixel groups to maintain a sharp edge boundary.

Abstract: A method for adjusting photographing settings of a digital camera through motion detection, for determine to adjust the photographing settings of the digital camera according to vibrations of the digital camera or shaking of a photographed object. The adjusting method includes the following steps. At least two continuous pre-photographed images are captured. A central area and an edge area are set at same positions in the pre-photographed images respectively. Feature points are found in the pre-photographed images. A local motion vector and a camera motion vector are respectively determined according to a relation of the feature points on the central area and the edge area. A subject motion vector is generated according to a difference between the local motion vector and the camera motion vector. Adjustment of the photographing settings of the digital camera is determined according to a maximum value in the subject motion vector and camera motion vector.

Abstract: An image processing device includes an image acquisition module, a memory module, and an image signal processing module, for performing an image enlargement and enhancement. The image acquisition module sequentially reads in an image block, including a unit pixel matrix and an exterior pixel matrix, wherein each pixel matrix includes a plurality of pixels and each pixel is associated with a parameter. The memory module stores a plurality of predefined edge patterns. The image signal processing module compares a loaded image block with predefined edge patterns, and determines if it is an edge block. Then, the image signal processing module further classifies its pixels into two groups, and calculates a continuous separating boundary between the two groups. Finally, the image signal processing module enlarges an edge block by placing new pixels inside its unit pixel matrix, wherein the new pixel parameters are extrapolated from the two classified pixel groups to maintain a sharp edge boundary.