Mobile Depth Sensing Using Spectral Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing depth sensing cameras on mobile devices face challenges such as interference from ambient light, declining precision with distance, noise, and light interference, making accurate object dimensioning difficult and requiring expensive components.
Innovation Solution
A hybrid device combining a digital camera and a depth sensing device, with a dimensioning processor that acquires digital images and depth data, segments shapes, associates segments with objects, and computes dimensions using depth data, filtering noise and correcting distortions to provide precise measurements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If depth sensing cameras use structured light, then depth measurement capability is achieved, but precision declines with distance and ambient light interference occurs
Solution Approach 1:
The patent segments the depth sensing process into multiple wavelength channels (e.g., 7 wavelengths from 730nm to 1060nm). By dividing the spectral range into discrete segments, the system can selectively measure and subtract ambient light contributions at each wavelength, thereby maintaining precision while operating in various lighting conditions.
Solution Approach 2:
The patent introduces spectral information as an intermediary medium between the structured light projection and depth calculation. By analyzing the spectral characteristics of reflected light across multiple wavelengths, the system can distinguish between projected structured light and ambient light, enabling precise depth measurement even in challenging lighting environments.
2Measurement precision
If depth sensing cameras use structured light, then depth data can be acquired, but expensive ASIC components are required for processing
Solution Approach 1:
The patent replaces complex dedicated ASIC hardware with a software-based spectral analysis approach using standard camera sensors. By utilizing the camera's existing spectral response characteristics and applying computational algorithms to analyze multi-wavelength images, the system achieves accurate depth measurement without requiring expensive application-specific integrated circuits.
Solution Approach 2:
The patent enables standard digital camera sensors to perform multiple functions: capturing visible light images for color information and simultaneously capturing spectral information for depth measurement. This multi-functionality eliminates the need for dedicated depth sensing hardware, reducing device complexity and cost while maintaining measurement accuracy.
3Measurement precision
If stereo cameras are used for dimensioning, then depth information can be obtained, but a significant baseline is required and low light degradation occurs
Solution Approach 1:
The patent transitions from spatial baseline separation (stereo vision requiring physical distance between cameras) to spectral dimension separation. By using multiple wavelength channels instead of multiple spatially separated sensors, the system achieves depth measurement capability with a single camera, eliminating the baseline requirement and improving low-light performance through spectral analysis.
Data Source
AI summary
A device and method of dimensioning using digital images and depth data is provided. The device includes a camera and a depth sensing device whose fields of view generally overlap. Segments of shapes belonging to an object identified in a digital image from the camera are identified. Based on respective depth data, from the depth sensing device, associated with each of the segments of the shapes belonging to the object, it is determined whether each of the segments is associated with a same shape belonging to the object. Once all the segments are processed to determine their respective associations with the shapes of the object in the digital image, dimensions of the object are computed based on the respective depth data and the respective associations of the shapes.