Training methods, equipment, and storage media for edge detection models of smart lockers
CN115661481BActive Publication Date: 2026-06-30BEIJING JINGHANG COMPUTING & COMM RES INST
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING JINGHANG COMPUTING & COMM RES INST
- Filing Date
- 2022-11-10
- Publication Date
- 2026-06-30
AI Technical Summary
Technical Problem
Existing methods for detecting openings in smart lockers have high hardware costs, poor maintainability, and are difficult to install, affecting the lifespan of the wiring.
Method used
By employing a lightweight convolutional neural network model and combining it with a low-cost area array industrial camera, an edge point detection model is constructed through image acquisition and stitching to achieve high-precision real-time detection of the opening position of the smart locker.
Benefits of technology
It reduces hardware costs, simplifies the maintenance process, achieves high-precision opening position detection, and does not increase computer hardware costs.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN115661481B_ABST
Abstract
This invention relates to a training method, device, and storage medium for an edge point detection model of a smart locker. The method includes the following steps: acquiring top views of the smart locker in different opening states; for each top view of the smart locker, extracting multiple sample images; obtaining the coordinates of the edge points at the opening of the smart locker marked in each sample image; constructing a sample dataset by combining the sample images and the corresponding edge point coordinates; constructing a lightweight convolutional neural network model; training the model based on the sample dataset; optimizing the model parameters using a gradient descent algorithm; and obtaining a trained smart locker opening edge point detection model.
Need to check novelty before this filing date? Find Prior Art