The patent's inventor is Chen, Maolin (
This patent was filed on
From the background information supplied by the inventors, news correspondents obtained the following quote: "One or more embodiments relate to a target analysis method, apparatus and computer-readable medium, and more particularly, to a target analysis method, apparatus and computer-readable medium based on a depth image and an intensity image, which may accurately detect each portion of a desired video target from a video stream, thereby obtaining motion information of the target.
"Along with various applications such as three-dimensional (3D) animation, games, human interface, and the like, there are wide research activities on video target analysis technology in current computer video and model identification regions. To perform an animation simulation, each portion of the video target may need to be detected in advance and motion information of the target may need to be obtained.
"To detect each portion of the video target, a bottom-up method and a top-down method may be used. In the bottom-up method, each portion of the target may be detected using characteristics of each portion of the target. For example, a skin color or a line type of arms and legs may be detected, and then a registration may be performed of detected candidate portions based on the detected information. The bottom-up method may significantly rely on typical characteristics of each portion of the target, and may be influenced by background noise. In the top-down method, a repeated retrieval may be performed on a space of the target shape, a hypothesis may be suggested, and a hypothesis evaluation may be performed. In the top-down method, a searching space may need to be significantly reduced to accelerate a calculation speed.
"In general, a foreground may need to be segmented from a source image in a video target analysis process. Specifically, in the foreground segmentation, the target may be accurately detected by deleting a complex background region. A conventional foreground segmentation method may be performed based on a color image of a charge-coupled device (CCD) camera. Since a background of the color image is significantly complex, a segmentation calculation amount based on a foreground of the color image may be significantly great and inaccurate. The foreground segmentation method is performed based on a current depth image, however, and may be inaccurately performed.
"Once the foreground segmentation is inaccurately performed, a consecutive target analysis subsequently performed may be significantly difficult to perform, and the target analysis result may be influenced by interference of background noise having similar characteristics. Also, rich edge characteristics may frequently occur in the complex background region, and the edge characteristics may become noise to influence an analysis result of the edge characteristic in the processes of detecting each portion of the target and of evaluating the hypothesis.
"Accordingly, there is a demand for a target analysis method, apparatus and computer-readable medium which may accurately segment a foreground to improve a target analysis performance."
Supplementing the background information on this patent, VerticalNews reporters also obtained the inventor's summary information for this patent: "One or more embodiments provide a target analysis method, apparatus and computer-readable medium which may accurately segment a foreground to improve a target analysis performance.
"One or more embodiments also provide a target analysis method and apparatus which may realize an accurate ridge verification.
"One or more embodiments also provide a target analysis method and apparatus which may use an intensity image and a depth image of a Time of Flight (TOF) without being limited to use of a color image.
"The foregoing and/or other aspects are achieved by providing a target analysis apparatus having a depth image and an intensity image of a target as an input, the target analysis apparatus including: a body detection unit to detect a body of the target from the intensity image of the target, a foreground segmentation unit to calculate an intensity threshold value in accordance with intensity values from the detected body, to transform the intensity image into a binary image using the intensity threshold value, and to mask the depth image of the target using the binary image as a mask to thereby obtain a masked depth image, and an active portion detection unit to detect an active portion of the body of the target from the masked depth image.
"The body detection unit may be trained using a statistical learning method to detect, from the intensity image, components constituting the body of the target, and to select, from a plurality of candidates for the components of the body, a candidate to meet a geometry constraint as the component of the body.
"The foreground segmentation unit may have as the intensity threshold value, a value obtained by averaging the intensity values from the detected body.
"The active portion detection unit may include a ridge detection unit to detect a plurality of ridges from the masked depth image, and a ridge verification unit to verify the plurality of ridges and to select an optical ridge.
"The ridge detection unit may divide a depth image of the mask into the active portion of the body and a non-active portion of the body, and calculate a plurality of ridge candidates of each of the ridges from a candidate of the active portion.
"The ridge detection unit may classify the depth image of the mask into the active portion and the non-active portion using an average depth value of the non-active portion as a threshold value.
"The ridge verification unit may determine a connection point between the active portion and the non-active portion in accordance with the body detected by the body detection unit, connect the determined connection point and an end of one of the plurality of ridges to generate an imaginary ridge, and verify the plurality of ridges and the imaginary ridge.
"The non-active portion of the body of the target includes a human head and torso, and the active portion includes human arms and legs.
"The foregoing and/or other aspects are achieved by providing a target analysis method based on a depth image and an intensity image of a target, the target analysis method including: detecting a body of the target from the intensity image of the target, calculating an intensity threshold value in accordance with intensity values from the detected body, transforming the intensity image into a binary image using the intensity threshold value, masking the depth image of the target using the binary image as a mask to thereby obtain a masked depth image, and detecting an active portion of the body of the target from the masked depth image.
"According to another aspect of one or more embodiments, there is provided at least one computer readable medium including computer readable instructions that control at least one processor to implement methods of one or more embodiments.
"Additional aspects of embodiments will be set forth in portion in the description which follows and, in portion, will be apparent from the description, or may be learned by practice of the disclosure."
For the URL and additional information on this patent, see: Chen, Maolin. Target Analysis Apparatus, Method and Computer-Readable Medium. U.S. Patent Number 8744168, filed
Keywords for this news article include:
Our reports deliver fact-based news of research and discoveries from around the world. Copyright 2014, NewsRx LLC
Most Popular Stories
- New Hershey's Logo Revealed
- Americans Still Pessimistic Despite Economic Growth
- Obama's Delay on Immigration Creates Uncertainty
- Startups Offer Smartphone Banking Apps
- Mexico's Pemex Forecasts 6.7% Drop in 2014 Crude Production
- Echeveste Steps Down, Perez Steps Up at VPE
- Clippers Deal Started With 2 Numbers
- 'Longmire' Cancelled, Looks for New Network
- Illinois Issues Fracking Rules
- Hip-Hop Takes Up Ferguson Cause