Retail Store Product Image Analysis with Edge-Based Redundancy Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Retail stores face challenges in efficiently analyzing large volumes of images for product inventory management due to redundant data, network bandwidth consumption, inaccurate image stitching, and low confidence in identifying new products or products with new packaging, leading to potential loss of sales opportunities.
Innovation Solution
Implementing edge computing to select a subset of images based on spatial and temporal contextual data, forming clusters of products with low confidence scores for annotation, and combining images using product recognition for accurate stitching to reduce redundancy and improve analysis efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all images are transmitted for analysis, then image analysis accuracy is improved, but network bandwidth consumption increases
Solution Approach 1:
The system segments the image processing workflow into two parts: initial filtering and selection performed at the edge device, and detailed analysis performed at the cloud server. This segmentation allows only necessary images to be transmitted, reducing network bandwidth consumption while maintaining analysis accuracy for relevant products.
Solution Approach 2:
The edge device performs preliminary actions by analyzing images locally and selecting only those containing new products or products with new packaging before transmission. This preliminary filtering reduces the volume of data transmitted over the network while ensuring that all potentially important images are captured for accurate analysis.
2Reliability
If redundant images are processed, then comprehensive product identification is improved, but processing time increases
Solution Approach 1:
The system extracts and removes redundant images from the processing pipeline by using the edge device to identify and filter out duplicate or unnecessary images before transmission. This extraction of redundant data maintains comprehensive product identification while significantly reducing processing time at the cloud server.
Solution Approach 2:
The system changes the parameter of image selection by using confidence scores and product novelty detection at the edge device. By adjusting the selection criteria to focus on images with low confidence scores or containing new products, the system processes only necessary images, reducing processing time while maintaining identification completeness.
3Speed
If image stitching is performed without product recognition, then stitching speed is improved, but stitching accuracy deteriorates
Solution Approach 1:
The system performs preliminary product recognition and key point identification before the stitching process. By pre-identifying products and their key points in each image, the stitching algorithm can quickly locate and align corresponding features, maintaining high stitching speed while achieving accurate results through product-aware alignment.
Solution Approach 2:
Product recognition acts as an intermediary between raw images and the stitching process. The identified products and their key points serve as intermediate features that guide the stitching algorithm, enabling both fast processing through pre-computed features and high accuracy through product-based alignment constraints.
Data Source
AI summary
In some aspects, an edge computing system may receive, a plurality of images. An image in the plurality of images may be associated with products in a retail store. The edge computing system may select a subset of images in the plurality of images based on spatial contextual data associated with each image in the plurality of images, a level of redundancy between images in the plurality of images, and temporal contextual data associated with each image in the plurality of images. The edge computing system may transmit, to a cloud computing system, the subset of images for image analysis.