Stacked Feature Extraction Element for On-Sensor CNN Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image feature extraction methods are time-consuming due to processing requirements for generating and transferring image data, which has not been adequately accelerated.
Innovation Solution
A feature extracting element with a light-receiving substrate and laminated substrates containing convolution and pooling processing units, along with a controlling unit to control convolution and sampling operations, allowing for efficient feature extraction by directly processing pixel values without data storage or transfer.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If feature extraction is performed by processing captured image data through generation and transfer, then feature extraction can be achieved, but processing time becomes excessively long
Solution Approach 1:
The patent merges the image capture function and feature extraction function into a single integrated device. The light-receiving substrate captures images while the laminated substrates perform feature extraction processing directly on the captured data, eliminating the need for separate data transfer and processing steps. This integration resolves the time loss caused by sequential processing in traditional systems.
Solution Approach 2:
The patent introduces a third dimension by stacking multiple functional substrates vertically. The light-receiving substrate is laminated with convolution processing substrates and pooling processing substrates, creating a three-dimensional integrated structure. This vertical integration enables parallel processing of feature extraction operations, dramatically improving processing speed while reducing the time required for feature extraction.
2Measurement precision
If traditional image processing methods are used, then complete image processing can be performed, but hardware resources and processing complexity increase
Solution Approach 1:
The patent extracts only the essential feature extraction functions from the complete image processing pipeline. By implementing only convolution processing and pooling processing on the laminated substrates, the system achieves effective feature extraction without the complexity of full image processing. This selective extraction of functions reduces hardware resource requirements while maintaining the necessary processing accuracy.
Solution Approach 2:
The patent segments the feature extraction process into distinct functional layers: light-receiving substrate for image capture, convolution processing substrate for feature detection, and pooling processing substrate for feature aggregation. Each substrate is optimized for its specific function, reducing overall device complexity while achieving high feature extraction accuracy through specialized processing at each stage.
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 significantly reduces processing time and hardware resources, enabling miniaturization and maintaining processing speed even with increased pixel counts, while facilitating high-accuracy feature extraction.
Implementation Method 1
a light-receiving substrate where a plurality of light-receiving elements for photoelectrically converting received light are two-dimensionally arrayed
Data Source
AI summary
An element includes a plurality of light-receiving elements to photoelectrically convert light received from an object, a convolution processing unit to perform convolution operation on signals that are output from the plurality of light receiving elements, and a pooling processing unit to sample a signal that is output from the convolution processing unit, based on a predetermined condition. The convolution operation of the convolution processing unit and the sampling of the pooling processing unit are repeated.