Camera CNN Feature Extraction Reduces SPAD Bandwidth
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing image processing systems using SPAD cameras face significant delays and high bandwidth demands due to the large data volume of target images being transmitted from hardware to software systems for processing, which is inefficient and time-consuming.
Innovation Solution
Integrating a preset target convolution layer with multiple convolution layers of a CNN within the camera to process and output target feature maps instead of raw images, reducing the data volume transmitted and thereby shortening transmission delays and saving bandwidth.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If the SPAD camera transmits the target image for large data volume, then the image data is complete and accurate, but the transmission delay is long and bandwidth demand is high
Solution Approach 1:
The patent extracts only the essential feature information from the complete target image through CNN processing. Instead of transmitting the entire high-volume image data, the system extracts key features (edges, textures, shapes) and transmits only these extracted features, which contain the essential information needed for object identification while dramatically reducing data volume and transmission delay
Solution Approach 2:
The patent introduces an intermediary processing system (CNN-based feature extraction module) between the SPAD camera and the object identification system. This intermediary processes the complete image data locally and transmits only the extracted feature representations, acting as a mediator that reduces transmission requirements while preserving essential information for accurate identification
2Loss of information
If the SPAD camera transmits the target image for large data volume, then the image data is complete and accurate, but the bandwidth demand is high
Solution Approach 1:
The patent extracts only the essential feature information from the complete target image through CNN processing. Instead of transmitting the entire high-volume image data, the system extracts key features (edges, textures, shapes) and transmits only these extracted features, which contain the essential information needed for object identification while dramatically reducing data volume and transmission delay
Solution Approach 2:
The patent segments the complete image data into essential features and non-essential data. Through CNN processing, the system identifies and separates the critical information elements (object boundaries, key textures, distinctive shapes) from redundant pixel data, transmitting only the segmented essential features to reduce overall transmission volume while maintaining identification accuracy
Data Source
AI summary
An image processing method, device, storage medium and camera are provided. The method, applied to the camera, comprises: capturing a target image; acquiring a target feature map of the target image through a preset target convolution layer, wherein the target convolution layer includes at least one of a plurality of convolution layers of a convolutional neural network (CNN); and outputting the target feature map. That is to say, after the target image is captured by the camera, the target feature map may be acquired by processing the target image through the target convolution layer pre-integrated in the camera. In this way, the camera transmits the target feature map only to reduce the transmitted data volume, thereby being capable of shortening transmission delay and saving bandwidth required by image transmission.