Thailand Zhenghe College (Bangkok) Electronic

Circuit Seminar and

University-Enterprise Cooperation Matching Meeting

Contents

1.Agenda

2.Conference Participating Units Information

2.1Huaiyin Institute of Technology
2.2 Rajamangala University of Technology Thanyaburi
2.3 Ubon Ratchathani Technical College
2.4 Chonburi Technical College
2.5 Udon Thani Vocational College
2.6 Peng Shen Technology (Thailand) Co., Ltd.
2.7 BYD Co., Ltd.
2.8 China Communications Technology Co., Ltd
2.9 Changzhou College of Information Technology
2.10 Nanjing Vocational College of Information Technology

3. Abstract and Keywords of Papers

泰国郑和学院（曼谷）电子电路研讨会暨校企合作对接会会议议程

一、时间

2025年5月10日，会期1天

二、地点

线上：腾讯会议

上午链接https://meeting.tencent.com/dm/naDsS5gac2EJ 下午链接https://meeting.tencent.com/dm/vI27TzaRsx11

三、组织单位

主办单位：泰国郑和学院（曼谷）

承办单位：淮阴工学院、常州信息职业技术学院、南京信息职业技术学院、泰国坦亚武里皇家理工大学、腾乌多姆技术学院、春武里技术学院、乌隆塔尼职业学院、鹏鼎控股集团、比亚迪汽车有限公司、中邮建技术有限公司、泰国江苏商会

四、参会人员

江苏省教育厅领导、中泰院校领导、泰国江苏商会领导、企业代表、中泰专家、学者，约100人。

ThailandZhngheColleBangkok）ElectronicCiruitSminarandUiversity-EnterpriseCooperationMathingMetng

议程安排

Agenda of

Thailand Zhenghe College (Bangkok) Electronic Circuit Seminar and University-Enterprise Cooperation Matching Meeting

I.Time

May10,2025,withadurationof1day

II.Venue

Online: Tencent Meeting MorningMeetingLink:https://meeting.tencent.com/dm/naDsS5gac2EJ AfternoonMeetingLink:https://meeting.tencent.com/dm/vI27TzaRsx1I

II.Organizing Institutions

Host Institution:Thailand Zhenghe College (Bangkok)

Organizers: Huaiyin Institute of Technology, Changzhou College of Information Technology, NanjingVocational Collge of Information Technology,Rajamangala University ofTechnology Thanyaburi,Detudom Technical College, Chonburi Technical College,Udonthani Vocational College,Avary Holding, BYD Auto Co., Ltd., China Post Construction,Jiangsu Chamber of Com merce in Thailand.

IV.Participants

Leaders fromJiangsu Provincial Department of Education,universityleadersfrom China and Thailand, executives of Jiangsu Chamber of Commerce in Thailand, enterprise representatives, Chinese and Thai experts & scholars, with approximately 100 participants in total.

V.Agenda Arrangement

HuaiyinIstituteofechnologyHT)stablishedi958isanapplication-rientedniversitymainly ocusinnninvrldnnatltu tustatnat frst-classapplication-rienteddergraduateuiversitisnJiangsuProvinceheuniversitycurrentlyhas threecamuseithcamus reafaproximatl200mharemrethan400fulm ste and over2000facultymembers.Theuniversityoffers73undergraduatemajorsandhas14professional master’s degree authorizationpoints.Three ofits disciplinesrank among the top 1% globally in the ESI ranking.Theioengieerindsiplinewaslistedinthe2024SofScienceWorld-ClassDisciplieRankingthas obtainedmorethan1600nationalauthorizednventionpatentsandhasrankedamongthetop50ntheChina UniversityPatentTransferRankingHYITis actively promoting internationalized educationIt hasestablishedfriendlycooperativerelationswithmore than60universitiesin22countriesaround theworld and Taiwan regionSo far, it has enrolld and cultivated more than 500undergraduate and postgraduate nterna tional studentsfromoutrisrecentearHIThasbeenrpeatedlyawarddthtitfAdvanc CollectiveforInternational Student EducationinJiangsuProvince”

Rajamangala University of Technology Thanyaburi

Rajamangala University of Technology Thanyaburi (abbreviated as: RMUTT) is a national university under the direct domination of theMinistry of Education ofThailand,whichis a high-levelandcomprehensiveuniversitywith integrating higher education and scientificresearch.Currently,it comprisesnine universities,consisting ofmore than ten campuses across the country.The main campus (RMUTT) is located in Thayaburi area,Patun Thani Province,Bangkok,Thailand,covering an area ofmore than 1,800 acres.RMUTT is the recommendation university of Foreign Supervision Network of theMinistry of Education of China,and the degree obtained can be definitely certified by,and share thepreferential policies such ashouseholdregistration and tax-freecarpurchasepolicyforreturnedstudents.

Theuniversity owns 117undergraduateprogramswith more than 60 master’s anddoctoral programs,consisting of12Faculties withmore than 26,000students and1,200facultiesandstaffs,includingtheFacultyofTechnical Education,the Faculty ofAgricultural Science and Technology,the FacultyofHomeEconomicsandTechnology,theFacultyofInformation Technology,theFaculty ofBusinessAdministration,theFacultyofScience andTechnology,theFacultyofEngineering,theFacultyofArt,theFaculty ofHumanities,theFacultyofArchitecture,theFacultyofTraditionalThai Medicine,and theFaculty ofNursing.Theuniversity is conveniently located about a 30-minute drivefromBangkokSuvarnabhumi International Airport, a20-minute drive from Don MueangInternational Airport and about a 40-minute drive from the downtownBangkok.

The campusis linedwith trees and flowerswith the SoutheastAsian style architecture andbeautiful scenery.TheFacultyhas abrandnewstudent apartment,providing a comfortable living environment and excellent servicesfor all students;thelibrary isrich incollection of books;indoor and outdoor sportsenues,gymsandother sportfacilities suchas tennishalls swimmingpools,basketball hallsbowlingalleys,football fields,etc.areall wellequipped.Banks,restaurants,conveniencestores andotherservicefacilitiesarewell completed,andunlimitednetworkcoverswholefiledof the campus.

Changzhou College of InforimationTechnology (CCIT) is affiliatedwith theIndustry and InformationTechnologyDepartment of Jiangsu.It is the onlyinformation-focusednational exemplaryvocational collegeinJiangsuProvince,and hasbeenselected as aconstructionunit for China'sHigh-Level HigherVocational Colleges and Specialties Initiative.Itis alsolisted amongthehigh-levelvocationalcollegesin JiangsuProvince,servingas a national modelforthe informatizationof vocational education and training.The school is located in Changzhou HigherVocational Parkwith beautiful scenery andmodern atmosphere, whichntratesteachingientificresearchainingocational sill appraisal and social services.It covers an area of1118 acres and has a total construction area of 421,800 square meters.There are 11627 full-timestudents,83full-tmeinternational studentand94faculty members.The college adheres to open education, focuses on improving international level,and carries out a wide range of international exchanges and cooperation.It has established long-term and stable cooperativerelationshipswithmore than40institutions and agenciesin more than20countriesandregionsintheworld,suchasThailand,the United Kingdom,Germany,Japan,South Korea,South Africa,and other countries and regions,and carries out in-depth and allaround international cooperationintheareas ofeducation,teaching,scientific research cooperation,academic exchanges,and studentscultivation. The college actively carries out overseas schooling,cooperates with Huawei,BYD and otherfamousinternational enterprises,explores the constructionof anoverseas branch operation modein which schools enterprises,socialorganizations andotherdiversifiedsubjectsoperate incollaboration.IthasestablishedoverseaseducationbasesinSouth Africahailandandother laces,cultivatingearlya tousandnt national technical talentswith a global perspective.

TheNanjingVocational CollegeofInformationTechnology（theCollege)isa state-wnedpublicnstitutiona highevligctionalstionwiCcharactristiatnaldehihcatinalstit highqualitycializedhigcatioalstitondntifdbythistyfductiohivl ocatiltevartfninxlglt in JiangsuProvinceTheCollges predecessorisNanjingRadiIndustrychoolwhichwas established953as t firstecdarpcializedsholflroniafefoungftlepublfCharhla ChenThnPiinxvillrthest havepadvitttlhanatladr ly skilled graduates for the community.

NJCITasbuiltafsuchasrsatudydrangJonCultivationffessionalMastn Inter-collJWrkhpanrg-rlatedehicalrvicenepiforelatdhal servicesrelandRoadrihavectdasbertftChAriccal EducationAlianceChinaASEAN"Thousand SchoolsHandsTogether Initiative",twoSin-ForeignCulturalExchange ProgramftMnstfdatideiftbaWkhllaavctd thIndustducationratirjctf tMnistyfducationIationalCheeTeachngReourc ConstructaFrihtinlatigitarjtdblt brandmajorsforintenational talent training inJanguProvince during the14thFiveYearPanh5Gtechnlogy andapplicationfviualiulatincmetitinandpracticalaininlaformwererecmmendedasinemationally influentialocatinaleducationquipmentfteMinistyfEducationtosrvemanyeationaleventsuchst WorldkillCometitonandICKSThecolletakesthadidevelopinaNatonalHighVcationalTeching QualityCertificationInexSystemforIationaludentstoservetenatioaltalenttainingandcertificatinf internationalstuentshighecationalcllgecollaborationwitenterprisewehavejntycultivatd international studentsignificantlyenhancing nternationalinfluence.

AccordingtotheewroudofthenationalDoubleHighLevelPlanNJCIThasbeenbuildingitselfint vocationalhlwithteabiltytcultivatedstgushedocationalergraduatandestablishingitelfasaeh markcollegeforthe high-qualitydevelopmentofnationalvocational education.

Chonburi Technical College currently has a total area of 94 rai, located at205Moo3,BanBueng-KlaengRoad,Km.2,NongChak Subdistrict,BanBuengDistrict,Chonburi Province.ChonburiTechnical Collegewas originally'Chonburi Carpentry School’(August 1, 1939-1955).Laterin1956,thenamewas changed to'ChonburiTechnical School'1939.OnAugust 1,1939,studentshad to stay at the WeavingSchool ortheformerWomen'sTechnicalSchool because the school buildingwasnot yet completed.Therewere37alumni ofthe1939 class.StudentswhohadcompletedPrimary4were accepted.The study periodwas 3years.Mr.Prasat...Makthapwas the acting headmaster in November

1939.When the new school buildingwas completed,the students moved to study at Chonburi Carpentry Schoolherental sitefAranyikawatTemple frest tmpl）has 2workshopshead teacher'sresidene juniortacheridencendjanitrssien9twasppvedtachicatnal icatelevelVcationalCertificateHigherVocationalCertificate）andsoonheschoolwasrenamedhon buriVocationalCollegeCampus.iaccordancewiththepolicyof theVocational EducationDepartment.

Intheventofthenamechangeat thatimetudentspuishadaroleinemandingthatitbechanged acollegeandchangeditsname toChonburiTechnicalCollege'fromJanuary11979tothepresent

In1982,itwas approved to teachat theVocational Certificatelevel(Vocational Certificate)

ChonburiTechnical Collgeprovidesteaching at theVocational CertificatelevelVocational Certifcate, Highercatnaetficatcatonaetfcatvesubjctsandachlorfhlogy tice,1 subject.There area totalof3067students and180teachers andeducationalpersonnel.

Formerlyknown asUbonRatchathaniTechnical College2,itwas establishedaccording to the announcement of the Ministry of Education by theDepartment ofVocational Education (now the Officeof theVocational EducationCommission)onJuly28,1997tosupport the government's policy toexpand basic educationfrom9yearsto12yearsandtodistributevocationaleducationinstitutions to communities far from the province.In 1998-1999, the administration building,temporary classroombuilding,tmporaryworkshopbuildingstoryadministrationbuilding,4storeyperman classroombuilding,4-storey permanentworkshopbuilding andotherbuildingswerecompletedand officially opened for teaching.In 2000, the college was announced to change its name fromUbon Ratchathani Technical College 2 toDecha Udom Technical College.The villagers who owned the land kindly donated 72 rai of land for the collge. The college began organizing the management system using participatory management,systematicmanagement,and management byobjective underbudget constraints.But wehavereceived assistance from otheragenciesboth overnment and private sectors,helpingwith equipment,materials andtools for teaching and learning and accelerat ing thedevelopmentof thequalityof students andpersonneluntilwehaveachieved tangibleresults, receiving certificatesofhonor from high-levelagencies,bothgovernment andprivate sectors,and the collegehasbeenevaluatedasanoutstandingeducationalinstitutionbytheDepartmentofVocational Education,etc.Currently,there are approximately 2,500 students.

Udon Thani Vocational CollegeWas establishedonJuly7,1938at the residence of theViceroy and the governor ofUdonThaniProvince,formerlyknownasUdonThaniWeaverSchool.1940 Opened to teach the Tailor Department at Udon Thani WeaverSchool1948 The weaving department wasdissolvedtojointhe sewing department.Establishedas thefirst-classwomen's techniciandepartment and changed itsname toUdonThaniKarnchangSchool.1950,was approvedbytheMinistry of Education to teach at the lower vocational level in addition toElementary vocational level and accepting students in Mathayom 3 to continue their studies.1976 Udon Thani Vocational School together with UdonThaniTechnical School and establishedas acollegeVocational Education UdonThani on October 1, 1976 with the division into 2 campuses, the Udon Thani Technical Campus. and Udon Thani Vocational Campus Open to teach advanced professional sentence (vocational certificate）in the field of home economics. Fabric and tailoring in the same year. 2001 Received cooperation from Big C (Udon Thani) Co., Ltd. in accepting vocational students in retail business Get a vocational training in anestablishment and getafulltime employee of theCharoensri GrandRoyal Hotel. Admissiontostudyat thevocationallevelinHotelManagement.2003OpenedVocationalCertificate Program(VocationalVocational Certificate)BE2545Revised2003OfferedVocational Certificate Program(VocationalVocational Certificate)2003and offeredVocationalVocational Education level bilateral system Retail BusinessManagement which cooperates with Big C (UdonThani) Co., Ltd.,CP7-ElevenCo.,Ltd.in recruiting full-time employees to study.2004Teaching at the vocational levelandvocationalvocationallevel,bilateral systemBakeryand teaching at thevocationallevel in TourismBusinessManagement,which is accredited to educational standards.from the Officefor Accreditation and Quality Assessment.

We are Zhen Ding Tech Group (ZDT),the world’s No.1 PCB producer since 2017.Peng Shen Technology （Thailand) Co.,Ltd. (PST) is a joint venture with Saha Group,one of Thailand’s leading industrial companies.

ZDTproducePCBs,printed circuitboards—whichare like the“brain”of electronic devices.Products such as smartphones,computers,andautomotiveallneedPCBsto function.

InThailand,we are collaborated with Saha Group to establishsmart andmodernfactories.Ourgoalistoimprove productionefficiencyandcreatemorejobopportunitiesfor Thaipeople.Our PST factories are located inthe Saha Group Industrial Park,Prachinburi,Kabinburi City.

8May,lastThursdaywehadcelebratedtherandopen ing ofourfirstfactory,Withthisstrongartership,were excitedtosupportthegrowthofThailand’stechnologyand PCB industry. Our aim is to become a leading global PCB manufacturer andbuild aworld-classsmartfactoryforthe ASEAN area.

At ZDT,we also value our employees as our key strength.Weinvestinnewtalentsbyofferingtrainingand development,including sending Thai employees to China and other countries togainnew skills and learnfromdifferent cultures.

BYD

BYDCo.tdhereinafterreerred toasBYD”）wasestablishedinNovember1994headquarteredin ShenzhenGuangdongvincesusiepansfoumajrdustriabilerailtransitrewabl energyndlctronicsItisamongthFortue0companieandlistedbothHongongandShnn StockExchanges.Asofnow,Dhasappliedformorethan48thousandpatents andobtainedmoretha0 thousand authorized patentsworldwide.After 30years of fast growth,the companyhas established over30 industrialparksworldwide and has playedasignificant rolenindustriesrelated tolectronics, automobiles, new energy and rail transit.From energy generation and storagetoits applicationsBYDs dedicated to provid ing zero-emission energy solutions.BYDis listed on the Hong Kong and Shenzhen Stock Exchanges, with revenue andmarketcapitalization each exceedingRMB100billion.

BYDThailandPlantlocated inRayongProvince'sWHA IdustrialParkisYD'sfirst verseasmanu facturingfailtvingasrdutnbasrriarivetrhilwillathls domesticmarket and export to otherASEAN countries.

Covering 948,000 square meters, the BYD automotive plant was completed within just 16 months from groundbreakingmracingenergysavingandcarbonreductionprinciplesfailtyfaturesfullyatmt ed machinery and employs eco-friendly production processes alongside advanced logistics management systemtnpassesfoucorattivmaufcturing stagestamn,wligpaintinnds bly,ensuringthprdutifhighqualit,standrdizedlricvhiclfrthhaimarktWithaxi mum annual productioncapacityof50000completevehicles,thplantmanufacturemodelsincludingt BYD Dolphin, BYDATTO 3, BYDSeal, and BYD SEALION 6.Aditionally, it possesses the capability to produce critical components such as batteries and powertrain systems.

PST Company Introduction

China CommunicationsTechnology Co.,Ltd,is awholly-owned susidiary ofChinaCommunications ServicesroupJanguCopanywitisteredcaitalfRM00millionouded958thcom nyhasconsistentlycontributedtonationalnformationtechnologydevelopmentItsbusinessscopeencom passes allfeldsofcommunicationsandnformationservicesoldingcomrehensivequalifcationsincom municatonsbroadcasting,mhanicalndltricalngineeringwerconstrutin,andrnwork servingmajor lnts suchasChinaTelcomChinaMobilChinanicomandChinaower,comany alsoocusndustrieudinanortatnroadastinwerdcationandwatonseva AdheringtthhilosophyfroadVisionBiDataHlisticlaingIngratedReourcesa Wisdomrandcritytvragfrtclassnralnratincapabiltmmuicatini ingandatdvilssinsiutinpianm companyprovides tailoredsolutionsfrominformationandcommunicationnfrastructure toplatformsand applications,empowering clientssmart development.

Thecompanyhas longbeenrecognizedasthe strongestincomprehensivestrengthforcommunication engineeringserviceswithinJiangsuProvinceandconsistentlyranksamongthetopenterprisesofitskind nationwideIthasundertakenthousandsofnationalkeyprojctsandiscommittedtocontinuousinnovation. AsthefirstChineseenterprise toreceivetheUnitedNationsWSISWinnerAwardandthefirstinthe industrytgainthNationalQualityGoldwardtcompanyasxtensivebuinesscoverageItsdomesti operationsspan31provinces（municipalitiesandautonomousrgions),whileitsintrnationalpresence extends toover20 countries acrossAsia,Africa,SouthAmericaand Europe.

Research on the concrete crack detection basedondepthlearning and UAVvision

YinshanYu\*XuTang,GuochenTan,JiawenBi,Wenkai Huang,MingjianDin Facultyfctronicformatinngineringuaiyinnstitutfchologyuain03i yuyinshan@163.com

AbstractIn this paper,we propose a crack detectionmethod based on deep learning and unmanned aerial vehicle (UAV) vision.The bridge crack dataset is collected by a UAV and a motion camera.Combining multiple deep learning models,the bridge crack detection and segmentation networkmodel isestablished to train andevaluate thecollected images.To improve thefeature extraction ability of themodel in channel and space,C3SE and C3CBAM are used toreplace allC3 modules in the backbone network structure of You Only Look Once-v5 (YOLO-v5) respectively. Finally,mage segmentation and parameter calculation arecarriedoutforbridgecracks.Compared with the experimental results,the model with squeeze-and-excitation(SE)and convolutional block attention module (CBAM) attention mechanism has improved the detection effect of multi-target and complexcracksResults presentedinthiswork prove the performance of themethodwebuildupand show the convenience of the subsequent actual measurement of bridge crackswithfast speed and high accuracy.

Keywordsconcrete crack; deep learning;unmanned aerial vehicle;attentionmechanism

Contemporary challenges of multispectral remote sensing for water quality monitoring: a comprehensive review

YinshanYu\*ingDingHaiyiBian,hmedN.dallaXuTangMingjianin FacultyfctronicformatinngineringuaiyinnstitutofTechnologyuaian03hi yuyinshan@163.com

AbstractAgainst thebackdropofescalating concerns about deterioratingwater quality,this comprehensivereviewpaper delvesinto thehistoricalevolution andcurrent state ofwater quality testing,emphasizing the groundbreaking application of multispectral remote sensing technology. Firstly,theworking principles andmonitoring processes underlyingmultispectralremote sensing are exploredIn addition,theapplicationofmultispectralremote sensingtechnologyforwaterquality monitoring is inroducedNext,various monitoring methods,including empirical, semimpirical, analytical,nuclide tracer,sensor technology,andoptical remote sensing,are comprehensively discussed.Further extend exploration to the diverse data sources employed inmultispectral remote sensinmassinatlitandrlmtnindatoundroardqp remotesensndataarchivingandcommrcildatarovidersFinallyantcipatdndsn offerinsights intothe future directions of this technology,considering itshistorical and contemporary contexts.In conclusion,synthesizes keyfindingsemphasizing the pivotal roleofmultispectral remote sensing in advancing waterquality testing and addressing contemporary challenges inwater quality monitoring.

Keywordswater quality testing;multispectral remote sensing;water monitor

River Floating Object Detection System Based onImprovedYOLOv10

YinshanYu\*JinsongWeiMingjianDingXuTangJiawenBiGuochenTanWenkai Huan FacultfctronicformatinngineringuaiyinnstitutofTechnologyuaian03hin yuyinshan@163.com

AbstractThis paper proposes a river floating object detection system based on an improved YOLOv10,aiming at addressing the target misdetection andmissing detection problems caused by limited computing power and complex river environments.Firstly,we design the lightweight C2F-EDWIBmodule toreplace the traditional C2Fmodulein thebackbonenetwork,effectively achieving a lightweight model. Secondly, the Bi-directional feature pyramid network (BiFPN) is utilizedas theneck instead of the path aggregationnetwork PAN),with integration of P2feature layertoenhancethedetectioncapabilityofsmall targetsAdditionallyweintroducethelightweigh triplet attentioninto thefaturemapoutputoftheNeckpart tomitigateperformancelos.Finallyby incorporating the minimum point distance based IoU (MPDIoU) lossfunction,alignment accuracy between predictedframes andrealframesismprovedforenhanced overall positioningperformance of our model. Experimental results demonstrate that the proposed model outperforms othermodels in terms of parameters,computation,and accuracy forthe detection of riverfloating objects.

Keywords-performance floating object;C2F-EDWIB;BiFPN;MPDIoU;triplet attention

Design oflightweightbridgedefect detection network based on feature collaborativeperception and multi-attention

RendongJiulongXuiayanWangLiyuZhuangHaiBinXiajuhangiuTangandJiaxinh FacultylectronicInformationEngineeringHuaiyinnstitutofTechnologyHuaianChina zly418@126.com

AbstractBridge defect detection isvery important in traffic safety monitoring andmaintenancemanagement.In ordertomake thedetectionmodelrunefficiently onthe equipment with limited computing power, this paperimproves theYOLOv10model toimprove thedetection accuracy and efficiency.Byoptimizing themodel structure,thebridgedefectscanbeidentifiedquicklyandaccu rately,which provides technical support forbridge maintenance.This paper designs the overall frameworkof CFAWD-YOLO,including threemainparts \because feature extractionbackbone network, featurefusionneck network and detection headThe backbone network uses animproved C2f FEMA module and a LAEDS spatial attentionmodule.Through progressive channel expansionand SPPF's 3x3 convolutionkernel design,the feature extraction efficiency isimproved.Theneck networkenhancesfeatureexpressionthroughmulti-scalefeaturefusionandEMAattentionattention mechanismanduses SCDown andCfCIBmodules to achieve effective transmission of features. The detection head part designs a lightweight v10Detect DP-Head structure,which combines DAMF-CAandchannel pruning strategytoachieveefficient target detection.Themodelistestedon three datasets of BDD-1234, DACL10K and BHMADD, and good detection results are obtained. Taking the BDD-1234 data set as an example,the designed CFAWD-YOLO framework reduces the modelsizefrom16.6MBto 2.8MB comparedwiththeoriginalYOLOv10model,whichisreduced by 83.1% In terms of calculationandparameterquantity,it decreasedfrom29.3GFLOPs and8.035 \mathbf{M} to 15.2 GFLOPs and 3.908{M} respectively,with a ecrease of 48.1% and 51.4% respectively, whilethedetectionaccuracyremained basicallyunchanged.In thispaper,thedetectionaccuracyand computational efficiency of theYOLOv10 model on computing power-constrained devices are improvedbylightweight improvement.The experimental results showthat the optimized model can realize therapid and acurate identificationofbridge defectsunder thepremiseofensuring high precision.Thisprovides anefficientandfeasible solutionforbridge safetymonitoringandmaintenance,andhasstrongpractical applicationvalue.

Keywordsbridge defect detecting; Feature fusion neck network; detector head; Backbone network

Road CrackSegmentation Algorithm

Based onMulti-Attention Fusion and Defect Correction

RendngJijuhangyanWangjiiangulngXungndJax FacultylectronicnformationEngineeringHuaiyinnstitutofTechnologyHuaianhin wxygxy@163.com

AbstractAiming at theproblemsofincompletesegmentationdetaillossandinaccurateaccu racy ofroad cracks in complex environment and unbalanced crack pixelratio,this paper proposes an improved U-Netroad cracksegmentationmethod basedonmultiple attentionfusion anddefect correction.Firstly,theAtrousResidual Convolution(ARC)moduleisintroducedinto theencoder part to expand the receptive field through the dilated convolution of different expansion rates, enhance the ability ofmulti-scale feature extraction,andimprove detection performance for large-scale cracks. Secondly,the Multi-Attention Fusion Module (MAFM) is embedded in the bridging stage,combining theEfficient ChannelAttention (ECA)module and the Convolutional Block Module (CBM),which strengthen the model's ability to focus on the crack region and effectively suppressthebackgroundnoiseinterference.Finally,thedefect correctionmoduleDCM)isn duced in the decoder stage.Ituses the deep supervisionmechanism and adaptiverefinement to process the featuremap detailsanddetect small cracksoptimizestherecovery of thecrackedge detailfeaturesandimprovesthedetectionaccuracyofsmallcracks andshortbranchesofcracksh algorithminthispaperhascarriedoutmanyexperimentsontheroad crackdatasetCFDCRACK500 andself-madeHYCrack.Theexperimentalresults showthat theF1 scoresof thealgorithm oneach dataset can reach 73.21% 84.52% and 81.74% respectively.Comparedwiththecurrentmainstream cracksentationodelthisarccuracyandetailnritfthodlrettwhca important applicationvalueand promotion prospectforcurrent andfuture road crack detectionand maintenance.

KeywordsSemantic segmentation;Defect correction;Road crack;Atrous residual convolution;Multi-attention fusion

Design of lightweightbridge defect detectionnetwork based on feature collaborative perception and multi-attention

RendonJiYungXiyanWangiyuhuangHaiBinijuhangXiTangandJixih FacultylectronicInformationEngineeringHuaiyinnstituteofTechnologyHuaianhina zly418@126.com

AbstractBridge defect detection is very important in traffic safety monitoring andmaintenancemanagement.Inordertomakethedetectionmodelrunefficientlyontheequipmentwithlimited computing power,this paperimproves theYOLOv10model toimprove thedetection accuracy and efficiency.Byoptimizing themodel structure,thebridgedefectscanbeidentifiedquicklyandaccu rately,which provides technicalsupportforbridgemaintenanceThispaper designstheoverall framework ofCFAWD-YOLO,including threemain parts:feature extraction backbonenetwork, feature fusion neck network and detection head.The backbone network uses an improved C2f FEMA module and a LAEDS spatial attention module.Through progressive channel expansion and SPPF's 3x3 convolution kernel design,the feature extraction efficiency isimproved.The neck network enhancesfeatureexpressionthroughmulti-scalefeaturefusionandEMAattentionattention mechanism,andusesSCDown and C2fCIBmodules toachieve effective transmissionof features. The detectionhead part designs a lightweightv10Detect DP-Head structure,which combines DAMF-CA and channel pruning strategy to achieveefficient target detectionThemodelis tested on three datasetsofBDD-1234DACL10K andBHMADD,and good detection results areobtained. Taking theBDD-1234 data set as an example,the designedCFAWD-YOLOframework reduces the model sizefrom 16.6\mathbf{MB} to 2.8\mathbf{MB} comparedwiththeoriginalYOLOv10model,whichisreduced by 83.1% In terms of calculation andparameterquantity,it decreasedfrom29.3GFLOPs and8.035 \mathbf{M} to15.2 GFLOPs and3.908M,respectively,with a decrease of 48.1% and 51.4% respectively, while thedetectionaccuracyremainedbasicallyunchanged.Inthispaper,thedetectionaccuracyand computational effciency of the YOLOv10 model on computing power-constrained devices are improvedbylightweightimprovement.Theexperimentalresults showthat theoptimizedmodel can realize therapid andaccurate identificationofbridge defectsunder thepremiseofensuring high precision.Thisprovidesanefficient andfeasible solutionforbridge safetymonitoring andmainte nance,and has strongpractical applicationvalue.

Keywordsbridge defect detecting;Feature fusion neck network; detector head;Backbone network

Road CrackSegmentationAlgorithm Based on Multi-Attention Fusion and Defect Correction

RendongJiijuhangiyanWangijQiiuTangulngXuugndJai FacultylectronicnformationEngineeringHuaiyinnstitutofTechnologyHuaianhi wxygxy@163.com

AbstractAiming at the problems of incomplete segmentation, detail loss and inaccurate accu racyofroadcracksincomplexnvironment andunbalancedcrackpixelratio,thispaper proposesan improved U-Net road crack segmentation method based on multiple attention fusion and defect correction.Firstly,theAtrousResidual Convolution(ARC)module isintroduced into the encoder part to expand the receptive field through the dilated convolution of diferent expansion rates, enhance the ability of multi-scale feature extraction, and improve detection performance for large-scale cracks.Secondly,theMulti-AttentionFusionModule MAFM)is embedded in thebridging stage,combining the Efficient ChannelAttention (ECA)module and the Convolutional Block Module(CBM),which strengthen themodel'sability tofocuson the crackregion andeffectively suppress thebackgroundnoiseinterference.Finally,the defect correctionmoduleDCM)isntro duced in the decoder stage.It uses the deep supervision mechanism and adaptive refinement to process the feature map details and detect small cracks, optimizes the recovery of the crack edge detail features,andmproves thedetection accuracyof smallcracks and shortbranchesofcracksThe algorithm in this paperhascarried out manyexperiments ontheroad crackdatasetCFD,CRACK500 and self-madeHYCrack.Theexperimentalresults show that theF1 scores of thealgorithmoneach dataset can reach 73.21% 84.52% and 81.74% respectively.Comparedwiththecurrentmainstream cracksmentationdelsthiaeraccuracyandetailntritfthodelretterwhicha important application value and promotion prospect forcurrent andfuture road crack detection and maintenance.

KeywordsSemantic segmentation;Defect correction;Road crack;Atrous residual convolution;Multi-attentionfusion

Simulation ofA Novel Electron Diffraction Reconstruction Framework

Jialiang Chen

SchoolfctronicEngineeringhangzhouCollgoflfmationTchnlogyhangzhou164hi chenjialiang@ccit.js.cn

AbstractThis paperintroduces a novel electron diffractionreconstructionframework aimed at addresnghlmitationsofcurrentmetodssuchaswaccuracylowrconstructioneedn high computational cost.The proposed framework utilizes a simulationmethodology that involves generating diffraction patterns from various sample orientations and positions, determining the posi tionofdiffractionspotsusingetaanglesandcreatingacomprehensive patterndatabase.Grainrin tation is determined using this database,while grain position is identified based on spot intensity, which correlates with the electron beam's covered volume on the grain.The framework was tested using simulated atternsandtheresultsdemonstrateitsreconstructioncapabilitieswitharepored accuracy rate from multiple tests.Future work includes refining the simulation to better match real-worldconditionsincorporatingeuraltworksordiffractionspotrcognitionandutilizin GPU acceleration to enhance calculation speed.This research presents a promising approach to improve the efficiency and accuracy of electron diffraction reconstruction.

Keywords-ElectronDiffraction;ReconstructionFramework;Simulation;

Reply Form

Reply Form