一种近红外卷积图像传感器结构设计
By moving convolution computation forward in the near-infrared image sensor and using macro-pixel unit arrays and readout circuits to perform analog domain convolution operations, the problems of high power consumption and high latency in traditional architectures are solved, realizing the integration of perception and computing, which is suitable for edge intelligence applications.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHANGHAI INSTITUTE OF TECHNICAL PHYSICS CHINESE ACADEMY OF SCIENCES
- Filing Date
- 2026-03-10
- Publication Date
- 2026-07-17
AI Technical Summary
In existing technologies, there are limitations in resource-constrained and power-sensitive fields. Existing technologies cannot effectively solve the problem of efficient feature extraction in complex near-infrared environments. Data transfer in traditional architectures leads to high power consumption and high latency issues, making it difficult to meet the real-time requirements of fast-moving scenarios or highly dynamic environments.
Design a near-infrared convolutional image sensor structure that moves the convolution calculation to the sensing end. The analog domain convolution operation is realized through a macro-pixel unit array and readout circuit. The photocurrent flow is controlled by static random access memory and the convolution kernel operation is completed through a metal wire interconnection structure, and the voltage signal is directly output.
It achieves the integration of perception and computing, eliminates data transfer bottlenecks, significantly reduces system power consumption and latency, improves real-time response capabilities, and is suitable for edge intelligence application scenarios.
Smart Images
Figure CN121815106B_ABST