Research on Privacy Boundaries in Machine-Vision Data Collection

Abstract

After machine vision evolves from simple recording to recognition, tracking, and inference, image collection in public places is no longer merely a question of camera installation. It has become an issue concerning the intensity of personal-information processing and the boundaries of personality rights. Based on a review of Chinese literature from the past three years and current normative documents, this paper concludes that public visibility is not equivalent to permission for continuous computation, and that deleting original images does not eliminate derivative risks. The paper proposes a judgment framework consisting of three dimensions: scenario sensitivity, identifiability, and purpose necessity. It also proposes strategies in five areas: impact assessment, minimization parameters, edge anonymization, permission auditing, and full-lifecycle accountability. The study argues that machine-vision collection must have a specific necessary purpose as a precondition and should use low-intrusion alternatives and edge-side reduction by default, so that a proportional balance can be achieved among security, efficiency, and personal dignity.