401 results about "Foreground detection" patented technology

Systems and methods for foreground analysis in real-time video include background subtraction and foreground detection, shadow removal, quick lighting change adaptation, static foreground region detection, foreground fragment reduction, and frame level change detection. Processes include background image extraction and foreground detection, integrating texture information of the background image and a current frame to remove false positive foreground areas resulting from lighting changes, integrating pixel intensity information by determining a cross-correlation of intensities between a current frame and the background image for each pixel in a foreground mask to remove image shadows. Static foreground region detection and fragment reduction are also included.

The invention relates to a fire smog detection method based on motion characteristics and the convolutional neural network. Through reading a video file, a first image is stored as an original image,and smog detection on each frame of the video is carried out; firstly, the original image is added to background update as reference, a background model is further established, secondly, a foregroundimage is extracted through a difference method, the foreground image is filtered through a dark channel threshold image to acquire candidate smog areas, lastly, a depth convolutional neural network model after training is loaded to automatically extract high-level characteristics of the candidate smog areas, and whether the candidate smog areas are smog areas is determined according to extracted characteristic vectors. The method is advantaged in that the channel prior knowledge is added to motion foreground detection, common interference is effectively filtered, environment adaptability of adetection method is improved, the convolutional neural network is used for carrying out characteristic extraction of smog images, and detection efficiency is substantially improved.

An integrated image processor implemented on a substrate is disclosed. An input interface is configured to receive pixel data from two or more images. A pixel handling processor disposed on the substrate is configured to convert the pixel data into depth and intensity pixel data. In some embodiments, a foreground detector processor disposed on the substrate is configured to classify pixels as background or not background. In some embodiments, a projection generator disposed on the substrate is configured to generate a projection in space of the depth and intensity pixel data.

The invention discloses a suspicious target detection tracking and recognition method based on dual-camera cooperation, and belongs to the technical field of video image processing. The method comprises the steps that a panoramic surveillance camera is utilized for collecting a panoramic image, the improved Gaussian mixture modeling method is adopted for carrying out foreground detection, basic parameters of moving targets are extracted, a Kalman filter is utilized for estimating a movement locus of a specific target, the specific target is recognized according to velocity analysis, the dual-camera cooperation strategy is adopted, a feature tracking camera is controlled to carry out feature tracking on the moving targets, a suspicious target is locked, the face of the suspicious target is detected, face recognition is carried out, face data are compared with a database, and an alarm is given if abnormities exist. According to the suspicious target detection tracking and recognition method, the dual-camera cooperation tracking surveillance strategy is adopted, defects of a single surveillance camera on a specific scene are overcome, and the added face recognition function can assist workers in identifying the specific target to a greater degree; in addition, the tracking algorithm adopted in the method is good in real-time performance, target recognition and judgment standards are simple and reliable, and the operation process is fast and accurate.

The invention discloses a static gesture recognition method based on a finger contour and decision-making trees. The method comprises the steps that a Kinect depth image is used as a data source at first, the approximate coordinates of the palm are positioned through the Kinect skeleton tracking function, and a square area containing the palm is cut out with the coordinates as a center; the self-adaptive adjacent value method is used for conducting foreground detection on the area, and the palm contour is detected after appropriate image morphology processing is conducted on the foreground image; a circumference sequence curve is used for conducting modeling on the palm contour, and the extreme point pair method is utilized for accurately distinguishing each finger contour and a wrist contour and building gesture feature sets; at last, the decision-making trees are used for training and recognizing the gesture feature sets with different finger numbers.

The invention discloses a crowd exceptional event detecting method based on an LBP (Local Binary Pattern) weighted social force model. The method comprises the following steps of: computing a light stream vector of a sampled point based on a block-matching method; extracting dynamic textures of the sampled point by a time-space domain local binary pattern, and performing spectral analysis of Fourier transform; computing social force of the sampled point based on LBP weighted social force model; performing histogram quantization on the social force and performing classification on the video sequence based on a support vector machine to detect the exceptional behavior. Through combination of light stream and LBP frequency spectrum, the method provided by the invention innovatively calculates the social force to detect the exceptional behavior of the crowd, thus avoiding background modeling, prospect detection and detection and track of target, improving robustness and reduce calculated amount, and being particularly fit for occassions with large density of the crowd and complicated environment.

The invention discloses a moving target detection method, which comprises the steps of background modeling and foreground detection, wherein the background modeling is performed by using a gaussian mixture model, and an operation of model updating is performed so as to obtain B gaussian distributions, for a pixel in a frame image to be detected, if in the B gaussian distributions subjected to sequencing, at least a gaussian distribution is matched with the current pixel value, the pixel is a background pixel, otherwise, and the pixel is determined as a foreground pixel. According to the invention, foreground noises can be effectively filtered so as to obtain an extremely clean background, thereby, an efficient denoising effect is kept, and the accuracy and completeness of foreground object detection are enhanced.

The invention discloses an identity identification method based on self-established sample libraries and composite characters in video monitoring. The method comprises the steps that firstly, preprocessing is conducted on an acquired video, foreground detection is conducted, so that moving object information is obtained, then face detection is conducted on the basis of the moving object information, the detected face is identified, if the currently detected face can not be identified, a user is inquired so as to identify the detected face, and the identified face is added to the face sample library; if the face can not be detected in foreground information, pedestrian detection is conducted, a detected pedestrian is traced, gait period detection is conducted on a traced pedestrian image sequence, features of detected gait information of a period are extracted and identified, if identification fails, gaits are classified in the same mode of user identification and added to the gait sample library. The identity identification method provides a solution for identity identification on the condition of a lack of training sample diversity or small samples.

A method for removing false foreground image content in a foreground detection process performed on a video sequence includes, for each current frame, comparing a feature value of each current pixel against a feature value of a corresponding pixel in a background model. The each current pixel is classified as belonging to one of a candidate foreground image and a background based on the comparing. A first classification image representing the candidate foreground image is generated using the current pixels classified as belonging to the candidate foreground image. The each pixel in the first classification image is classified as belonging to one of a foreground image and a false foreground image using a previously trained classifier. A modified classification image is generated for representing the foreground image using the pixels classified as belonging to the foreground image while the pixels classified as belonging to the false foreground image are removed.

An object detection method and an object detection system, suitable for detecting moving object information of a video stream having a plurality of images, are provided. The method performs a moving object foreground detection on each of the images, so as to obtain a first foreground detection image comprising a plurality of moving objects. The method also performs a texture object foreground detection on each of the images, so as to obtain a second foreground detection image comprising a plurality of texture objects. The moving objects in the first foreground detection image and the texture objects in the second foreground detection image are selected and filtered, and then the remaining moving objects or texture objects after the filtering are output as real moving object information.

The invention provides a detection method of parking against rules. The detection method of the parking against the rules comprises the following steps: appointing a forbidden parking area of each frame of image in a collected video sequence; conducting a prospect detection to detect a goal of the prospect; following the detected goal to judge whether a target enters into the forbidden parking area, if yes, painting a color histogram within the range of the forbidden parking area, if no, conducting the detection continuously;and monitoring the duration of the color histogram after the color histogram changes to judge whether the duration is longer than the preset time, if yes, judging that a car against the rules parks in the forbidden parking area, if no, no car against the rules parks in the forbidden parking area. The invention further provides a detection device of parking against the rules. The detection device comprises an appointing module, a prospect detection module, a following module, a painting module and a judging module. The detection method and the device of parking against the rules can effectively achieve the automatic detection of the parking against the rules.

Systems and methods for foreground analysis in real-time video include background subtraction and foreground detection, shadow removal, quick lighting change adaptation, static foreground region detection, foreground fragment reduction, and frame level change detection. Processes include background image extraction and foreground detection, integrating texture information of the background image and a current frame to remove false positive foreground areas resulting from lighting changes, integrating pixel intensity information by determining a cross-correlation of intensities between a current frame and the background image for each pixel in a foreground mask to remove image shadows. Static foreground region detection and fragment reduction are also included.

The invention discloses a method of judging a shielding state of a pick-up lens based on a video image signal. The method is characterized by conducing image processing for a video image by taking of a camera, extracting background of the video, obtaining foreground, conducting binarization processing for the foreground, dividing a foreground block, according to the size of a pixel area of the foreground block, obtaining a foreground detection unit, filtering a nuisance area in the foreground detection unit, obtaining a suspicious sheltering area, and if the sheltering area value is judged to be larger than preset sheltering threshold value, the camera is confirmed to be sheltered. The method is used for solving the technical problem that the camera is sheltered so as to cause monitoring failure. The method has the advantages of being simple in method, high in precision, and wide in adaptability.

A temporal sequence of images is acquired of a dynamic scene. Spatial gradients are determined using filters. By taking advantage of the sparseness of outputs of the filters, an intrinsic background image is generated as median filtered gradients. The intrinsic background image is then divided into the original sequence of images to yield intrinsic foreground images. The intrinsic foreground images can be thresholded to obtain a detection mask.

The invention provides a smoke detection method based on a support vector machine. The method comprises the following steps of: (101) establishing background images: partitioning a continuous frame image into W*H blocks and then establishing background image for each subblock; (102) acquiring a foreground detection area; (103) calculating the characteristics of the foreground detection area; and (104) recognizing smoke areas according to the characteristics by adopting the support vector machine. The invention also provides a smoke detection device based on the support vector machine. The smoke detection method and the device based on the support vector machine can effectively detect smoke which slowly changes so as to solve the practical problem of recognition difficulty of smoke slowly changed in large indoor scenes, such as warehouses, large laboratories, and the like.

A system and method for analyzing the motions of an object based on the silhouettes of the object are provided. The system includes a foreground detector, a contour extractor, a model generator, a corner histogram generator, and a value of similarity measuring unit. The foreground detector detects a moving foreground object from an input image. The contour extractor extracts silhouette contour of the detected foreground object, and the model generator generates mean value histogram models as references to determine motions of the object. The corner histogram generator generates corner histograms of hierarchical multiband in the extracted contour signal, and the value of similarity measuring unit calculates a value of similarity between the generated corner histogram of a current frame and the average model histogram in a histogram unit, measures a value making a value of similarity with the calculated current frame histogram maximum, and determines the measured value as a posture of the object in the current frame.

The present invention describes a system for recognizing objects from color images by detecting features of interest, classifying them according to previous objects' features that the system has been trained on, and finally drawing a boundary around them to separate each object from others in the image. Furthermore, local feature detection algorithms are applied to color images, outliers are removed, and resulting feature descriptors are clustered to achieve effective object recognition. Additionally, the present invention describes a system for extracting foreground objects and the correct rejection of the background from an image of a scene. Importantly, the present invention allows for changes to the camera viewpoint or lighting between training and test time. The system uses a supervised-learning algorithm and produces blobs of foreground objects that a recognition algorithm can then use for object detection / recognition.

A method for foreground detection using multi-modality fusion graph cut includes the following steps. A video frame of a video sequence is inputted. A foreground region in the video frame is designated using adaptive background Gaussian Mixture Model and a threshold value. The foreground region is segmented using a multi-modality fusion graph cut procedure. A computer program product using the method and a system for foreground detection using multi-modality fusion graph cut are also disclosed herein.

An object tracking system which includes a plurality of camera devices, each of which captures image data, a plurality of thin-client processors, each of which is connected to a respective camera device via a local connection, each thin-client processor hosting a thin-client application that processes the captured image data to obtain two-dimensional foreground region information by using a background subtraction algorithm, and a server hosting an object tracking application that receives the foreground region information from each thin-client processor via a network and generates a three-dimensional visual hull corresponding to each foreground region represented in the received foreground region information, wherein the object tracking application generates identification and position data corresponding to each three-dimensional visual hull. The thin-client application uses two-dimensional object tracking to identify each object in the foreground region information, and sends each object identity to the object tracking application with the foreground region information.

The invention provides an automatic labeling method for a human joint based on a monocular video. The automatic labeling method comprises the following steps: detecting a foreground and storing the foreground as an area of interest; confirming an area of a human body, cutting a part and obtaining an outline of sketch; obtaining a skeleton of the human body and obtaining a key point of the skeleton; utilizing a relative position of face and hands to roughly estimate the gesture of human body; and automatically labeling the point of human joint. During an automatic labeling process, the sketch outline information, skin color information and skeleton information from sketch of human body are comprehensively utilized, so that the accuracy for extracting the joint point is ensured. According to the automatic labeling method provided by the invention, the accurate and efficient cutting for the part of human body is performed, the gesture information of each limb part is obtained and a beneficial condition is supplied to the next operation for obtaining and treating feature vectors of human body.

The invention discloses a regional average value kernel density estimation-based moving target detecting method in a dynamic scene. The method comprises the following steps of: firstly, initializing a background model; secondly, building a time and space background model for describing the dynamic complex scene by using a training sample in a background modelling process and considering the time sequence characteristics of pixel points in a video frame and the space characteristics in the adjacent regions of the pixel points; thirdly, continuously updating the background model by using the new video frame sample in a moving target detecting process; fourthly, adapting to the instantaneous background change by the regional kernel density estimating method and adapting to the continuous background change by using single Gauss background model, wherein the combination of the two models can fast and accurately adapt to the continuous change of the background and increases the executing efficiency of the method at the same time; and finally performing a foreground detecting method by providing an adjacent region information amount-based method so as to further remove noise points and inanition of a moving target in the background region in the detecting process and more completely extract the moving object in the foreground. The method can be widely applied to alarming the suspicious moving target in an intelligent monitoring system in an outdoor scene or a prohibited military zone and has wide market prospect and application value.

The invention discloses a detecting, positioning and corresponding method of a multi visual angle moving target, and belongs to the video monitoring and control technology field. The detecting, positioning and corresponding method comprises the following procedures that: a foreground detection is performed to the video frequency image of various visual angles, a two-valued foreground image is obtained; a space field model is established according to the two-valued foreground image, the three-dimensional reconstruction is performed in the space field model, and the three-dimensional reconstruction result of a moving target is obtained; the analysis is performed to the three-dimensional reconstruction result, the moving target is detected and positioned in the space field, and the space location of the moving target is obtained; the projection is performed to various visual angles according to the space location of the moving target, and the coincidence relation of the moving target among various visual angles is confirmed. The invention is characterized in that the screen treatment capacity is strong, the operational speed is quick, and the real time requirement of the video monitoring and control can be met.

The invention relates to a self-calibration method for a pickup camera based on an image and motion information of a moving target in a video, which comprises: carrying out prospect detection for the video containing the moving target, and extracting moving target regions; extracting characteristics from each moving target region; roughly classifying the moving target regions; extracting mutually vertical three vanishing points from the images and the motion information of the massive moving target regions; and combining height information of the pickup camera to finish full calibration of the pickup camera for monitoring a scene. The method replaces workload and error of manual calibration, and is used for obtaining an actual point distance of a three-dimensional world through point distance in the images and obtaining an actual line included angle of the three-dimensional world through a line included angle in the images based on measurement of the images or the video, used for monitoring object classification and recognition in the scene and compensating inherent perspective deformation of two-dimensional image characteristics, and used for monitoring object recognition in the scene based on a three-dimensional model to obtain a three-dimensional posture and a track and effectively help a system to comprehend behaviors in the scene.

An image surveillance system and a method of detecting whether an object is left behind or taken away are provided. The image surveillance system includes: a foreground detecting unit which detects a foreground region based on a pixel information difference between a background image and a current input image; a still region detecting unit which detects a candidate still region by clustering foreground pixels of the foreground region, and determines whether the candidate still region is a falsely detected still region or a true still region; and an object detecting unit which determines whether an object is left behind or taken away, based on edge information about the true still region.

The invention provides a real-time high-precision people stream counting method, which comprises two steps of moving target foreground extraction and foreground area pedestrian detection, wherein the step of moving target foreground extraction comprises the following steps: carrying out foreground detection on an obtained video frame sequence to obtain a foreground area comprising moving targets, such as pedestrians, vehicles and the like; and the step of foreground area pedestrian detection comprises the following steps: carrying out pedestrian detection on the foreground area by utilizing an off-line training deformable component model to determine the amount and the positions of the pedestrians in the foreground area, and tracking the subsequent movement of pedestrian targets by taking a current frame detection result as start to judge and record a situation that people streams enter and leave a gate. The method extracts the foreground area which comprises a target on the basis of a background subtraction method, so that an algorithm satisfies a real-time calculation condition, the pedestrian detection based on the deformable component model guarantees the high precision of people stream counting, a real-time people stream counting method with high precision and good shielding resistance is provided, and the real-time high-precision people stream counting method exhibits high practical value and good development prospect.

The invention discloses a method for detecting and tracking a night running vehicle. The method is implemented according to the following steps of: (1) foreground detection: selecting vehicle lamp brightness to carry out foreground detection, and detecting a vehicle lamp region through detecting whether the brightness of each frame of pixel point in a video stream is larger than a set threshold or not; (2) noise elimination: removing most noise points from a binary image obtained from the step (1) and more accurately obtaining a foreground target; (3) vehicle lamp matching: pairing two vehicle lamps according to a corresponding principle, finding out a big front lamp pair and representing the vehicle by using the big front lamp pair; (4) vehicle lamp pair tracking: after finishing the vehicle lamp pairing according to the steps, tracking the vehicle lamp so as to realize the tracking of the vehicle; and (5) after retracking the target, matching the vehicle lamp pairs and finally obtaining a vehicle to be detected. The method for detecting and tracking the night running vehicle, disclosed by the invention, has the following advantages that: the night vehicle is detected by using the vehicle lamp characteristic; and in the method, simplicity for extracting algorithm characteristic and stable vehicle detecting effect are obtained.

A system is disclosed that can find an image of a foreground object in a still image or video image. Finding the image of the foreground object can be used to reduce errors and reduce the time needed when creating morphs of an image. One implementation uses the detection of the image of the foreground object to create virtual camera movement, which is the illusion that a camera is moving around a scene that is frozen in time.