基于AI视觉识别的日用品模具表面缺陷检测数据

通过工业相机采集日用品模具表面图像，利用深度学习算法识别划痕、凹陷、毛刺等7类典型缺陷的技术，融合YOLOv5目标检测与ResNet34分类网络，实现对塑料模具表面多维缺陷的自动化识别。本缺陷检测数据具有以下应用场景：在企业内部，1. 通过实时采集模具表面缺陷数据，构建注塑工艺参数与缺陷类型的关联模型，实现工艺参数动态调整。2.可持续监测模具缺陷演变规律，建立磨损生命周期预测模型。在企业外部，1. 积累的数万条缺陷样本数据库，为行业协会制定《塑料模具表面缺陷分级标准》提供数据支撑。2. 与机器视觉厂商合作开发专用检测设备，集成本缺陷识别模型后，使同类企业检测效率大大提升。首先，图像采集：采用2000万像素工业相机，配合环形LED光源多角度照射。其次，特征提取：通过CNN卷积层提取缺陷纹理、几何等特征。再次，量化模型，构建缺陷指数公式：P=w1*S+w2*D+w3*C+w4*T+w5*L，其中，w1至w5为各个输入量参数的权重，其大小值利用Halcon深度学习框架构建缺陷检测模型时，通过反向传播算法自动调整大小。S为缺陷面积比，D为最大深度，C为颜色对比度差异（相邻阶灰度差≥5%时C为0.2，相邻阶灰度差≥10%时C为0.5，相邻阶灰度差≥15%时C为0.7，否则C为0），T为缺陷类型系数（划痕T=0.7、毛刺T=0.5、裂纹T=0.9，否则T为0），L为缺陷位置影响因子（边缘L=0.8，中心L=0.5，否则L为0）。最后，根据计算得到的P值对日用品模具表面缺陷进行分级，优级(P≤0.28)，良级(0.28<P≤0.44)，差级(P>0.44)。

This technology collects surface images of daily necessities molds via industrial cameras, uses deep learning algorithms to identify 7 typical defects including scratches, dents and burrs, and integrates YOLOv5 object detection and ResNet34 classification networks to realize automated recognition of multi-dimensional defects on plastic mold surfaces. This defect detection dataset has the following application scenarios:
Internal enterprise applications:
1. Construct a correlation model between injection molding process parameters and defect types by real-time collection of mold surface defect data, to achieve dynamic adjustment of process parameters.
2. Sustainably monitor the evolution law of mold defects and establish a wear life prediction model.
External enterprise applications:
1. The accumulated tens of thousands of defect sample databases provide data support for industry associations to formulate the "Classification Standard for Surface Defects of Plastic Molds".
2. Cooperate with machine vision manufacturers to develop dedicated testing equipment; after integrating this defect recognition model, the detection efficiency of similar enterprises will be greatly improved.
The dataset construction process includes the following steps:
First, image collection: A 20-megapixel industrial camera is used, paired with a ring-shaped LED light source for multi-angle illumination.
Second, feature extraction: Defect features such as texture and geometry are extracted through CNN convolutional layers.
Third, quantitative modeling: A defect index formula is constructed as $P = w_1 imes S + w_2 imes D + w_3 imes C + w_4 imes T + w_5 imes L$, where $w_1$ to $w_5$ are the weights of each input parameter, which are automatically adjusted via backpropagation algorithm when building a defect detection model using the Halcon deep learning framework. Specifically, $S$ denotes the defect area ratio; $D$ denotes the maximum depth; $C$ denotes the color contrast difference (C=0.2 when the adjacent gray level difference ≥5%, C=0.5 when ≥10%, C=0.7 when ≥15%, otherwise C=0); $T$ denotes the defect type coefficient (T=0.7 for scratches, T=0.5 for burrs, T=0.9 for cracks, otherwise T=0); $L$ denotes the defect location impact factor (L=0.8 for edge positions, L=0.5 for central positions, otherwise L=0). Finally, surface defects of daily necessities molds are graded based on the calculated P value: Excellent grade (P ≤ 0.28), Good grade (0.28 < P ≤ 0.44), Poor grade (P > 0.44).

搜集汇总

背景与挑战

背景概述

该数据集是基于AI视觉识别的日用品模具表面缺陷检测数据，包含10439条记录，每日更新，用于识别7类典型缺陷并进行缺陷分级，应用于企业内部工艺调整和外部标准制定及设备开发。