Depth-Based Touch Detection Using Surface-Relative Distance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current depth-based touch detection methods are prone to errors and false-positive detections due to their reliance on fixed camera setups, lack of normalization of pixel depth-data, and the use of predefined thresholds, which limits their applicability to dynamic scenes and arbitrary surfaces.
Innovation Solution
A depth-based touch detection method that generates a depth map, identifies and normalizes a region indicative of an object relative to a surface, and applies a classifier to determine touch events, allowing for dynamic camera positions and arbitrary surface shapes, with optional temporal filtering for improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional depth-based touch detection methods are used with fixed camera setups and predefined thresholds, then the system can detect touch events, but the measurement precision deteriorates due to errors and false-positive detections
Solution Approach 1:
The patent transforms the depth detection parameters from camera-relative coordinates to surface-relative coordinates by introducing a reference plane defined by three surface points. This parameter transformation allows the system to measure distances relative to the actual surface geometry rather than the camera position, eliminating false positives caused by camera movement and improving measurement precision for arbitrary surface shapes
Solution Approach 2:
The system performs preliminary identification and characterization of the surface geometry before conducting touch detection. By first establishing a reference plane through three identified surface points and computing surface normals, the system prepares the geometric framework needed for accurate subsequent touch measurements, ensuring reliable detection even on arbitrary surfaces
2Adaptability or versatility
If predefined thresholds are used to distinguish touch and no-touch conditions, then the classification process is simplified, but the adaptability deteriorates because the system cannot handle dynamic scenes or arbitrary surfaces
Solution Approach 1:
The patent implements a dynamic reference frame that adapts to each scene by identifying surface points and computing a reference plane specific to the detected surface geometry. Rather than using a fixed camera-coordinate system, the reference plane dynamically adjusts to match arbitrary surface shapes and orientations, enabling the system to handle diverse scenarios without requiring predefined thresholds for each surface type
Solution Approach 2:
The reference plane acts as an intermediary coordinate system between the camera and the arbitrary surface. By introducing this intermediate geometric construct defined by three surface points, the system mediates the transformation from unstructured depth data to structured touch detection coordinates, simplifying the classification process while maintaining high adaptability to different surfaces
3Ease of manufacture
If flood fill methods are used to segment the finger and detect touches, then the implementation is straightforward, but the measurement precision deteriorates due to lack of depth normalization
Solution Approach 1:
The patent applies parameter transformation by converting raw depth values into distances relative to the computed reference plane. Each depth value is normalized by subtracting the reference plane distance at that pixel location, transforming the measurement from camera-relative to surface-relative coordinates. This parameter change maintains implementation simplicity while dramatically improving depth measurement accuracy
Data Source
AI summary
Systems, methods, and computer readable media to improve the operation of detecting contact between a finger or other object and a surface are described. In general, techniques disclosed herein utilize a depth map to identify an object and a surface, and a classifier to determine when the object is touching the surface. Unlike the prior art, a measure of the object's “distance” is made relative to the surface and not the camera(s) thereby providing some measure of invariance with respect to camera pose. The object-surface distance measure can be used to construct an identifier or “feature vector” that, when applied to a classifier, generates an output indicative of whether the object is touching the surface. The classifier may be based on machine learning and can be trained off-line before run-time operations are commenced. In some embodiments, temporal filtering may be used to improve surface detection operations.