摘要:
Methods and a system for identifying reflective surfaces in a scene are provided herein. The system may include a sensing device configured to capture a scene. The system may further include a storage device configured to store three-dimensional positions of at least some of the objects in the scene. The system may further include a computer processor configured to attempt to obtain a reflective surface representation for one or more candidate surfaces selected from the surfaces in the scene. In a case that the attempted obtaining is successful, computer processor is further configured to determine that the candidate reflective surface is indeed a reflective surface defined by the obtained surface representation. According to some embodiments of the present invention, in a case the attempted calculation is unsuccessful, determining that the recognized portion of the object is an object that is independent of the stored objects.
摘要:
Structured light active sensing systems transmit and receive spatial codes to generate depth maps. Spatial codes can't be repeated within a disparity range if they are to be uniquely identified. This results in large numbers of codes for single transmitter/single receiver systems, because reflected ray traces from two object locations may be focused onto the same location of the receiver sensor, making it impossible to determine which object location reflected the code. However, the original code location may be uniquely identified because ray traces from the two object locations that focus onto the same location of the first receiver sensor may focus onto different locations on the second receiver sensor. Described herein are active sensing systems and methods that use two receivers to uniquely identify original code positions and allow for greater code reuse.
摘要:
The subject disclosure is directed towards detecting planes in a scene using depth data of a scene image, based upon a relationship between pixel depths, row height and two constants. Samples of a depth image are processed to fit values for the constants to a plane formulation to determine which samples indicate a plane. A reference plane may be determined from those samples that indicate a plane, with pixels in the depth image processed to determine each pixel's relationship to the plane based on the pixel's depth, location and associated fitted values, e.g., below the plane, on the plane or above the plane.
摘要:
According to the invention, a computer-implemented method for processing digital stereo image content, comprises the steps of estimating a perceived disparity of the stereo image content and processing the stereo image content, based on the estimated perceived disparity.
摘要:
A powerful, scalable, and reconfigurable image processing system and method for processing stereo image data is described. A general purpose, reconfigurable engine with toroidal topology, distributed memory, and wide bandwidth I/O is described for solving real applications at real-time speeds. The reconfigurable image processing system can be optimized to efficiently perform specialized computations, such as real-time video and audio processing. This reconfigurable image processing system provides high performance via high computational density, high memory bandwidth, and high I/O bandwidth. Generally, the reconfigurable image processing system and its control structure include a homogeneous array of 16 field programmable gate arrays (FPGA) and 16 static random access memories (SRAM) arranged in a partial torus configuration (Fig. 46). The reconfigurable image processing system also includes a PCI bus interface chip, a clock control chip, and a datapath chip. It can be implemented in a single board. It receives data from its external environment, computes correspondence, and uses the results of the correspondence computations for various post-processing, industrial applications. The reconfigurable image processing system determines correspondence by using non-parametric local transforms followed by correlation. These non-parametric local transforms include the census, and rank transforms. Other embodiments involve a combination of correspondence, rectification, a left-right consistency check, and the application of an interest operator.
摘要:
A method and system for registering data by first acquiring the data from a scene by a sensor at different viewpoints, and extracting, from the data, three- dimensional (3D) points and 3D lines, and descriptors associated with the 3D points and the 3D lines. A first set of primitives represented in a first coordinate system of the sensor is selected, wherein the first set of primitives includes at least three 3D points. A second set of primitives represented in a second coordinate system is selected, wherein the second set of primitives includes any combination of 3D points and 3D lines to obtain at least three primitives. Then, using the first set of primitives and the second set of primitives, the 3D points are registered with each other, and the 3D points with the 3D lines to obtain registered primitives, wherein the registered primitives are used in a simultaneous localization and mapping (SLAM) system.
摘要:
Systems and method for generating depth maps using active sensing technology, for scenes with moving objects, is disclosed. One aspect provides for a method that includes estimating areas in adjacent frames that correspond to a moving object by generating a probability map for each received frame, the probability map comprising a probability value at each pixel. The method also includes computing a convex temporal average map using a plurality of the reflected structured light frames including at least the prior frame received at time t-1, the received frame received at time t, and the next frame received at time t+1, the value at each pixel of the convex temporal average map weighted and normalized by the probability map at each pixel at each time;. The method also includes determining the codewords at each pixel in the convex temporal average map, and generating a depth map from the determined codewords.
摘要:
Methods and apparatus for making environmental measurements are described. In some embodiments different devices are used to capture environmental information at different times, rates and/or resolutions. Environmental information, e.g., depth information, from multiples sources captured using a variety of devices is processed and combined. Some environmental information is captured during an event. Such information is combined, in some embodiments, with environmental information that was captured prior to the event. Environmental depth model is generated in some embodiments by combining, e.g., reconciling, depth information from at least two different sources including: i) depth information obtained from a static map, ii) depth information obtained from images captured by light field cameras, and iii) depth information obtained from images captured by stereoscopic camera pairs. The reconciliation process may involve a variety of information weighting operations taking into consideration the advantages of different depth information sources and the availability of such information.