人工智能在制造业应用探索ExplorationofIndustrialAIAIHistory人工智能的过去人工智能发展的两个主要阶段TwomainstagesofAI物理符号系统PSSH(PhysicalSymbolSystemHypothesis)左侧:赫伯特·西蒙右侧:艾伦·纽厄尔Left:HerbertA.SimonRight:AlanNewell2012年以前:专家系统Before2012:ExpertsDefinitionMemoryReceptorControlEffectMovement32012年以后:深度学习After2012:DeepLearning4人工智能发展的两个主要阶段TwomainstagesofAI人工智能技术AITechnologies自然语言理解NaturalLanguageProcessing计算机视觉ComputerVision5人工智能三要素ThreeElementsofAI反馈Feedback数据3.Data模型2.Model丹方Prescription灵材Materials人工Artificial智能Intelligence深度学习计算框架DLcomputingframework计算力1.Computingpower人工智能大脑AIBrain6活的人工智能让产品更强大LiveAImakeproductsbetter7深度学习训练DeepLeaningTraining现象1Phenomenon1结果1Result1Phenomenon2Result2Phenomenon3Result3Phenomenon4Result4............Phenomenon1,000,000Result1,000,000现象Phenomenon1,000,001结果Result?8ThechallengesofAI人工智能的挑战ArtificialIntelligenceAI大量的人工高昂的成本AlotofmanualeffortsHighlaborcosts10准确率的提升ImprovementOfAccuracy准确率Accuracy成本Cost100%的准确率难度极大Theaccuracyof100%ImpossiblyDifficult11IndustrialAI智能制造传统工业视觉vs人工智能TraditionalIndustrialVisionvsAI传统工业视觉TraditionalIndustrialVision简单外观检测Simpleappearancedetection测距Distancemeasurement13传统工业视觉缺陷TraditionalIndustrialVisionLimitations固定位置(Fixedposition)固定环境(Fixedsetting)固定算法(Fixedalgorithm)14拍摄条件要求低(Lowshootingrequirements)异常扩展性好(Goodexpansibilityinexception)人工智能优势ArtificialIntelligenceAdvantages传统工业视觉vs人工智能TraditionalIndustrialVisionvsAI人工智能能做什么?WhatcanAIdo?分类-什么Classification:What?PlantIdentificationxAIDeepLearning形色10000Speciesx98%x30million15植物识别x人工智能检测-哪里Detection:Where?自动批改数学作业x人工智能CorrectmathhomeworkxAIDeepLearning98%x9millionx8000Years爱作业16人工智能能做什么?WhatcanAIdo?质检=什么+哪里QualityAssurance=What+Where分类Classification检测Detection+17人工智能能做什么?WhatAIcando?智能制造应用案例CasesinIndustrialAI零件视觉索引管理VisualIndexManagementforParts配件管理(PartsManagement)安全生产(SafetyControl)18NOYes配置检测ConfigurationInspection19智能制造应用案例CasesinIndustrialAI复杂场景类裸眼质量检测AutomationQAinComplexScenes20智能制造应用案例CasesinIndustrialAI曲面检测CurvedSurfaceMeasurement面临的挑战Challenges:划痕(Nicks)灰尘(Dust)多角度(Multi-angle)21智能制造应用案例CasesinIndustrialAI