TOF Camera Noise Reduction via Oxide Semiconductor and Deep Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current TOF cameras face challenges in achieving high temporal resolution and accurate distance measurement due to noise from leakage in transistors using single crystal silicon, which affects signal intensity and resolution, and require frequent measurements to compensate, leading to integration and downsizing issues in semiconductor circuits.
Innovation Solution
The use of deep learning for image processing in TOF systems, specifically with a U-Net method, to enhance distance image data accuracy by overlapping spatial slices and employing transistors with oxide semiconductor layers to reduce noise and leakage, allowing for higher temporal resolution and downsized chip design.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If transistors using single crystal silicon are used in TOF cameras, then manufacturing precision and electrical conductivity are improved, but noise from leakage increases and temporal resolution deteriorates
Solution Approach 1:
The patent changes the material parameter of the transistor from single crystal silicon to oxide semiconductor, which fundamentally alters the electrical characteristics. This material substitution reduces leakage current by several orders of magnitude while maintaining adequate conductivity when needed, directly resolving the noise and reliability issues without sacrificing manufacturing precision
Solution Approach 2:
The patent employs a hybrid structure combining oxide semiconductor transistors with photodetector elements in the imaging device. This composite approach leverages the low-leakage properties of oxide semiconductors for charge storage while maintaining sensitivity in the detection portion, achieving both low noise and high signal-to-noise ratio
2Speed
If exposure time is shortened to improve temporal resolution, then temporal resolution is improved, but signal intensity decreases
Solution Approach 1:
The patent extracts and separates the charge accumulation function from the detection function by introducing dedicated charge storage nodes and oxide semiconductor transistors. This allows the photodetector to capture photons during short exposure intervals while the separated storage elements accumulate and integrate the charge signals over extended periods, thereby maintaining both high temporal resolution and sufficient signal intensity
Solution Approach 2:
The patent implements preliminary charge accumulation in dedicated storage nodes before final readout. The oxide semiconductor transistors pre-charge and hold the accumulated charge from multiple short exposure periods, preparing the signal in advance for high-speed readout. This preliminary accumulation action enables temporal resolution improvement without sacrificing signal intensity
3Measurement precision
If measurements are repeated many times to improve resolution, then measurement precision is improved, but productivity decreases
Solution Approach 1:
The patent implements continuous charge accumulation across multiple measurement cycles using oxide semiconductor transistors with extremely low leakage. Instead of discrete repeated measurements with full readout cycles, the system continuously integrates charge from successive light pulses into the storage nodes, maintaining precision improvement while enabling continuous high-speed operation and大幅提升 productivity
4Volume of moving object
If chip size is reduced for integration, then device complexity is reduced, but manufacturing precision requirements increase
Solution Approach 1:
The oxide semiconductor transistor structure serves multiple functions simultaneously: it acts as both the switching element and the charge storage element, and provides both low-leakage operation and adequate conductivity. This multi-functionality reduces the total number of components needed on the chip, enabling size reduction without increasing fabrication precision requirements
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
This approach enables the acquisition of high-accuracy distance images with reduced noise and increased temporal resolution, facilitating the integration and miniaturization of semiconductor circuits in TOF cameras, suitable for applications like autonomous driving.
Implementation Method 1
a light source emitting light to an object; a light-receiving region receiving light reflected by the object and generating charges
Data Source
AI summary
An object is to obtain accurate distance image data by denoising. Another object is to realize distance image data acquisition in a short time by reducing the frequency of accumulating. A distance image processing system including a solid-state imaging element that can be used for three-dimensionally recognizing an object is provided for the utilization of autonomous driving of passenger cars, for example. Image processing including distance information obtained by a TOF system solid-state imaging element, a so-called TOF camera, is performed by utilizing deep learning. A high-accurate distance image with noise reduced by deep learning can be obtained.