Depth Sensing Using Active Coherent Signals and Occlusion Constraints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional sparse recovery processes are inadequate for exploiting the significant structure in scenes for coherent depth sensing, leading to ambiguities and poor reconstruction performance in depth mapping.
Innovation Solution
A model-based Compressive Sampling Matching Pursuit (CoSaMP) process is employed, enforcing occlusion constraints to recover depth maps, ensuring that only one reflector is identified per direction, and using a union of subspaces model to improve reconstruction accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional sparse recovery processes are used for depth sensing, then the system is simple to implement, but reconstruction performance is poor and ambiguities remain
Solution Approach 1:
The patent applies preliminary action by enforcing occlusion constraints during the reconstruction process. Before final depth map generation, the algorithm pre-identifies which reflectors should be visible or occluded based on geometric relationships, then uses these pre-determined constraints to guide the sparse recovery process. This preliminary structuring of the problem significantly improves depth sensing accuracy while keeping the implementation manageable through structured constraint enforcement.
2Reliability
If occlusion constraints are enforced to improve depth map accuracy, then reconstruction performance improves, but processing complexity increases
Solution Approach 1:
The patent applies segmentation by dividing the scene into discrete grid points and processing each direction independently through the occlusion constraint enforcement. The reconstruction algorithm segments the problem into manageable directional components, applying occlusion logic to each ray direction separately. This segmentation approach improves reliability by ensuring consistent occlusion handling while controlling processing complexity through modular, direction-by-direction computation.
Solution Approach 2:
The patent applies dimensionality change by transforming the depth sensing problem from a purely spatial reconstruction task into a constrained optimization problem in a higher-dimensional space that includes occlusion relationships. By adding the occlusion constraint dimension to the reconstruction process, the algorithm achieves more reliable depth maps while managing complexity through structured constraint formulation rather than brute-force processing.
3Adaptability or versatility
If multiple reflectors are present in the scene, then the scene complexity increases, but conventional methods produce ambiguities in depth reconstruction
Solution Approach 1:
The patent applies feedback by using the reconstructed depth map to inform subsequent reconstruction iterations. The algorithm reconstructs the depth map, then uses this reconstruction to update occlusion constraints, which feed back into the next reconstruction iteration. This feedback loop progressively resolves ambiguities caused by multiple reflectors, improving adaptability to complex scenes while reducing information loss through iterative refinement based on accumulated depth knowledge.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The approach significantly enhances depth sensing by reducing ambiguities and improving reconstruction performance, especially in scenarios with multiple reflectors, providing clearer and more accurate depth maps compared to conventional sparse recovery methods.
Implementation Method 1
The ability to acquire the reflected waveform allows coherent arrays to measure the time-of-flight of the transmitted pulse from the instance it is transmitted until the reflected pulse is received.
Data Source
AI summary
Reflectors in a scene are reconstructed using reflected signals. First, signals are transmitted to the scene, and the reflected signals are received by receivers are arranged in an array. The reflected signals are then processed to reconstruct the reflectors in the scene, wherein the processing enforces a model that a reflectivity of the scene in front and behind any reflector is equal to zero.