Sweeping robot obstacle identification method and system based on convolutional neural network

By analyzing image orientation gradients and edge transitions using convolutional neural networks, obstacle regions are identified and separated, solving the problem of inaccurate obstacle recognition in existing technologies and achieving clear obstacle determination in complex environments.

CN122265969APending Publication Date: 2026-06-23湖南鹏耀科技有限公司 +1
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-03-12
Publication Date
2026-06-23

AI Technical Summary

Technical Problem

Existing obstacle recognition methods struggle to maintain contour consistency in complex environments, making it difficult to identify boundary interruption locations. Target and background regions are prone to interference, and structural connections cannot be fully expressed, impacting the reliability of path planning.

Method used

By employing a convolutional neural network-based approach, edge transitions and boundary defects are identified through image orientation gradient region analysis. The channel activation mode of densely responding regions is adjusted to construct image response focusing regions and separate independent obstacle regions.

Benefits of technology

It achieves a clear and separable state of image structure under complex texture conditions, improves the reliability and completeness of obstacle spatial boundary determination, and enhances the accuracy of obstacle recognition.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122265969A_ABST
    Figure CN122265969A_ABST
Patent Text Reader

Abstract

The application relates to the field of environmental perception technology, in particular to a sweeping robot obstacle identification method and system based on a convolutional neural network, which comprises collecting an environmental image in a working direction and extracting scale level features, constructing a direction gradient region based on pixel transverse and longitudinal response changes, identifying a trend change along a continuous response path to form an edge jump region, analyzing a multi-direction structure connection state to position a boundary missing region, enhancing local response at a broken trend and limiting a receptive field range, and separating independent image regions according to texture contour closure relations. In the application, the stability expression of a structure boundary in a complex scene is realized through joint modeling of pixel response continuity and direction consistency, the contour integrity and region independence are guaranteed through the trend correlation and focus enhancement of the interrupted boundary, and the accuracy of obstacle boundary determination and the reliability of environmental perception are improved.
Need to check novelty before this filing date? Find Prior Art