基于Mobilenetv1-YOLOv4的低算力汽車數(shù)據(jù)脫敏技術(shù)研究

打開文本圖片集

中圖分類號(hào)：U462.3 收稿日期：2025-05-12

DOI：10.19999/j.cnki.1004-0226.2025.08.004

Research on Low-computational-power Vehicle Data Desensitization Technology Based on Mobilenetv1-YOLOv4

Zhu ZhanglinglLiu Peng2Wang Yixuan2 Wang Houzheng1Liu Yixuan2He Ting1 1.Anhui Pusi Standard Technology Co.，Ltd.，Wuhu 241ooo，China 2.CATARC Automotive Test Center（Wuhan）Co.，Ltd.，Wuhan 430056，China

Abstract：Withtherapiddevelopmentofintellgentandconnectedvehicles，anincreasingnumberofvehiclesareequippedwith advanceddriverassistancesystems thatrelyonextealsensorssuchascamerasandlidarforroadenvironmentperceptionanddriving decision-making.isprocssivolvesalageamountofextealinteractieifoationtransmsioaingeicledatauityissuesincreasinglyprominent.Toensuredriingsafety，extealpersonalinformationthatcanotbecollectedwithpersonalcosetand transmitedovertheetworkuchasicenseplatesandacsustbeonyiedncludingdeletingimagesthacontainiiable naturalpersosoaplingpartialcontouringproesingtoacialifomationineimags.Tddresstisisue，tispaperrosesa low-computational-powervehicledata desensitization methodbasedonMobilenetv1-YOLOv4.This methodcombinessuper-resolution enhancementreconstructiontechnologyforimageenhancement，enablingdesensitizationofboth externalfacesandvehiclelicense platesatlowomputatioalcosts.Iomplexenvionments，teodelcanstllmaitainhighdetectionacuracyandeal-tiepefor mance.Theimplementationof tisresearchnotonlycomplieswiththerequirementsofthe“SeveralProvisionsontheManagementof AutomotiveDataSecurity”andthenatioalstandardGB/T4464-2024“GeneralRequirementsforAutomotiveData”，butalsohasimportant practical significance for protecting personal privacy and enhancing the security of automotive data.

Key words：Object detection; YOLOv4;Image enhancement;Anonymization processing

1前言

隨著智能網(wǎng)聯(lián)汽車的飛速發(fā)展和高階自動(dòng)駕駛技術(shù)的廣泛應(yīng)用，車輛已不再僅是一個(gè)交通工具，更是一個(gè)移動(dòng)的數(shù)據(jù)中心，涉及大量的數(shù)據(jù)采集與傳輸。（剩余9774字）