本页翻译不是最新的。点击此处可查看最新英文版本。

自动化视觉检查

使用异常检测和定位方法自动执行质量控制任务

自动化视觉检查 (AVI) 是一套用于对于制造产品缺陷进行检测和分类的方法。现代视觉检查使用机器学习和深度学习算法来辅助和改进质量保证流程。

自定义缺陷检测必须准确地确定给定图像中缺陷的位置，并对缺陷类别进行分类。一般情况下，您可以使用最先进的有监督深度学习模型来检测不同大小的缺陷，例如使用最新的 you only look once (YOLO) 算法。性能最优的模型能够实时表征和定位缺陷。

您可以使用异常检测深度学习方法来确定所制造产品的图像是正常还是异常。此外，您还可以使用异常定位生成精确且可解释的结果。您可以通过这种方法使用异常图来创建缺陷的可视化表示。

具体选择哪个检测模型来自动化视觉检查任务取决于若干因素。这些因素包括可用于正常和异常样本的训练数据量、要识别的异常类的数量以及理解和监控预测值所需的定位信息的类型。有关详细信息，请参阅Getting Started with Anomaly Detection Using Deep Learning。

要执行自动化视觉检查，请从附加功能资源管理器下载 Automated Visual Inspection Library for Computer Vision Toolbox™。有关下载附加功能的详细信息，请参阅获取和管理附加功能。某些功能还需要 Deep Learning Toolbox™。

函数

全部展开

groundTruthGround truth label data
imageDatastore图像数据的数据存储
objectInsertionDatastoreCreate synthetic labeled datastore for training an instance segmentation or object detection network (自 R2025a 起)
sceneLabelTrainingDataCreate training data for scene classification from ground truth (自 R2022b 起)
splitAnomalyDataSplit data into training, validation and testing sets for anomaly detection (自 R2023a 起)
insertObjectInImageInsert object from source image into image (自 R2025a 起)
objectInsertionDatastoreCreate synthetic labeled datastore for training an instance segmentation or object detection network (自 R2025a 起)
trainEfficientADAnomalyDetectorTrain EfficientAD anomaly detection network (自 R2024b 起)
trainFCDDAnomalyDetectorTrain fully convolutional data description (FCDD) anomaly detection network (自 R2022b 起)
trainFastFlowAnomalyDetectorTrain FastFlow anomaly detection network (自 R2023a 起)
trainPatchCoreAnomalyDetectorTrain PatchCore anomaly detection network (自 R2023a 起)
anomalyThresholdOptimal anomaly threshold for set of anomaly scores and corresponding labels (自 R2022b 起)
efficientADAnomalyDetectorDetect anomalies using EfficientAD network (自 R2024b 起)
fcddAnomalyDetectorDetect anomalies using fully convolutional data description (FCDD) network for anomaly detection (自 R2022b 起)
fastFlowAnomalyDetectorDetect anomalies using FastFlow network (自 R2023a 起)
patchCoreAnomalyDetectorDetect anomalies using PatchCore network (自 R2023a 起)
classifyClassify image as normal or anomalous (自 R2022b 起)
predictPredict unnormalized anomaly scores (自 R2022b 起)
anomalyMapPredict per-pixel anomaly score map (自 R2022b 起)
anomalyMapOverlayOverlay heatmap on image using per-pixel anomaly scores (自 R2022b 起)
percentileNormalizerCreate percentile normalizer for anomaly map normalization (自 R2024b 起)
normalizeNormalize anomaly score map using percentiles (自 R2024b 起)
viewAnomalyDetectionResultsView anomaly detection results (自 R2022b 起)
evaluateAnomalyDetectionEvaluate anomaly detection results against ground truth (自 R2022b 起)
anomalyDetectionMetricsAnomaly detection metrics (自 R2022b 起)
yoloxObjectDetectorDetect objects using YOLOX object detector (自 R2023b 起)
trainYOLOXObjectDetectorTrain YOLOX object detector (自 R2023b 起)
detectDetect objects using YOLOX object detector (自 R2023b 起)
counTRObjectCounterCount objects using CounTR deep learning model (自 R2025a 起)
countObjectsCount objects using CounTR deep learning model (自 R2025a 起)
densityMapDensity map of object count generated using CounTR deep learning model (自 R2025a 起)
caliperMeasure distance in image data using caliper tool (自 R2025a 起)
uicaliperInteractively measure distance in images using a caliper tool (自 R2025a 起)

主题

精选示例

Count Objects Using CounTR Model

Count Objects Using CounTR Model

Demonstrates few-shot object counting using the Counting Transformer (CounTR) model, which counts objects in an image by identifying a few exemplars without additional training.

Perform Metrology Edge Measurements and Alignment

Perform Metrology Edge Measurements and Alignment

Perform distance measurements of fill lines in soda bottles using a caliper tool, and develop a classification model for normal and anomalous fill line images.

Detect Small Objects Using Tiled Training of YOLOX Network

Detect Small Objects Using Tiled Training of YOLOX Network

Detect small objects in full-resolution images using tiled training of a you only look once version X (YOLOX) deep learning network.

Detect Defects Using Tiled Training of EfficientAD Anomaly Detector

Detect Defects Using Tiled Training of EfficientAD Anomaly Detector

Detect and localize defects on anomalous chewing gum images by training an EfficientAD anomaly detection network on tiled normal images.

Detect Defects on Printed Circuit Boards Using YOLOX Network

Detect Defects on Printed Circuit Boards Using YOLOX Network

Detect, localize, and classify defects in printed circuit boards (PCBs) using a you only look once version X (YOLOX) deep learning network.

Identify Defects in Air Compressors Using Spectrogram Images

Identify Defects in Air Compressors Using Spectrogram Images

Detect and localize defects in acoustic recordings of air compressors using Mel spectrogram images and an EfficientAD anomaly detector.

Localize Industrial Defects Using PatchCore Anomaly Detector

Localize Industrial Defects Using PatchCore Anomaly Detector

Perform localization of anomalous defects in printed circuit boards (PCBs) using anomaly heat maps generated with the PatchCore anomaly detector.

Classify Defects on Wafer Maps Using Deep Learning

Classify Defects on Wafer Maps Using Deep Learning

Classify manufacturing defects on wafer maps using a simple convolutional neural network (CNN).

Detect Anomalies in Pills During Live Image Acquisition

Detect Anomalies in Pills During Live Image Acquisition

Detect anomalies in pills during live image acquisition.

(Image Acquisition Toolbox)

Detect Image Anomalies Using Explainable FCDD Network

Detect Image Anomalies Using Explainable FCDD Network

Use an anomaly detector to distinguish between normal pills and pills with anomalous chips or contamination.

Detect Image Anomalies Using Pretrained ResNet-18 Feature Embeddings

Detect Image Anomalies Using Pretrained ResNet-18 Feature Embeddings

Train a similarity-based anomaly detector using one-class learning of feature embeddings extracted from a pretrained ResNet-18 convolutional neural network.