基于改进Faster RCNN的马克杯缺陷检测方法

李东洁; 李若昊

doi:doi:10.3788/LOP57.041515

激光与光电子学进展, 2020, 57 (4): 041515, 网络出版: 2020-02-20

基于改进Faster RCNN的马克杯缺陷检测方法下载： 1395次

Mug Defect Detection Method Based on Improved Faster RCNN

李东洁 ^*李若昊

作者单位

哈尔滨理工大学自动化学院, 黑龙江哈尔滨 150080

图 & 表

图 1. FPN+RPN算法网络结构图

Fig. 1. FPN+RPN algorithm network structure diagram

下载图片查看原文

图 2. 部分训练样本。(a)包含四个缺陷,即缺口、划痕、两个斑点;(b)包含一种缺陷,即斑点;(c)包含两个缺口缺陷

Fig. 2. Partial training samples. (a) With four defects, one gap, one scratch, and two speckles; (b) with one speckle defect; (c) with two gap defects

下载图片查看原文

图 3. 部分测试样本。(a)包含一个缺口缺陷;(b)包含两个缺陷,即缺口和斑点;(c)包含一个划痕缺陷

Fig. 3. Partial test samples. (a) With one gap defect; (b) with two defects, one gap and one speckle; (c) with one scratch defect

下载图片查看原文

图 4. 基于ZF网络的训练损失。(a) RPN第一阶段训练损失;(b) Faster RCNN第一阶段训练损失;(c) RPN第二阶段训练损失;(d) Faster RCNN第二阶段训练损失

Fig. 4. Training loss based on ZF network. (a) Stage-1 training loss of RPN; (b) stage-1 training loss of Faster RCNN; (c) stage-2 training loss of RPN; (d) stage-2 training loss of Faster RCNN

下载图片查看原文

图 5. 改进后ZF网络的训练损失。(a) RPN第一阶段训练损失;(b) Faster RCNN第一阶段训练损失;(c) RPN第二阶段训练损失;(d) Faster RCNN第二阶段训练损失

Fig. 5. Training loss based on improved ZF network. (a) Stage-1 training loss of RPN; (b) stage-1 training loss of Faster RCNN; (c) stage-2 training loss of RPN; (d) stage-2 training loss of Faster RCNN

下载图片查看原文

图 6. 使用labelImg标注数据集

Fig. 6. Labeling data set with labelImg

下载图片查看原文

图 7. 马克杯缺陷检测结果对比图。(a)原始Faster RCNN检测结果;(b)加入FPN后的Faster RCNN检测结果

Fig. 7. Comparison of mug defect inspection results. (a) Original Faster RCNN; (b) Faster RCNN after FPN addition

下载图片查看原文

表 1改进前后的ZF网络结构

Table1. ZF network structure before and after improvement

Type of layers		Number of convolution kernels		Step size
Before	After	Before	After	Before	After
Conv_1/1	Conv_1/2	96	64	7×7/2	3×3/1
Max pooling/1	Max pooling/1	96	64	3×3/2	2×2/2
Conv_2/1	Conv_2/2	256	128	5×5/2	3×3/1
Max pooling/1	Max pooling/1	256	128	3×3/2	2×2/2
Conv_3/3	Conv_3/2	384/384/256	256	3×3/1	3×3/1
Max pooling/1	Max pooling/1	256	256	3×3/2	2×2/2
	Conv_4/3		384		3×3/1
	Max pooling/1		384		2×2/2
	Conv_5/3		512		3×3/1

查看原文

表 2不同网络结构的分类性能

Table2. Comparison of various network structures on classification performance

Network structure	AP /%			Average detectiontime /s
Network structure		Scratch	Speckle	Gap
Faster RCNN	85.32	0	97.03	0.094
Faster RCNN and two layers of FPN	87.86	81.11	98.30	0.123
Faster RCNN and three layers of FPN	88.32	82.36	98.51	0.135
Faster RCNN and four layers of FPN	88.56	82.98	98.76	0.149

查看原文

李东洁, 李若昊. 基于改进Faster RCNN的马克杯缺陷检测方法[J]. 激光与光电子学进展, 2020, 57(4): 041515. Dongjie Li, Ruohao Li. Mug Defect Detection Method Based on Improved Faster RCNN[J]. Laser & Optoelectronics Progress, 2020, 57(4): 041515.

基于改进Faster RCNN的马克杯缺陷检测方法下载： 1395次

图 1. FPN+RPN算法网络结构图

Fig. 1. FPN+RPN algorithm network structure diagram

图 2. 部分训练样本。(a)包含四个缺陷,即缺口、划痕、两个斑点;(b)包含一种缺陷,即斑点;(c)包含两个缺口缺陷

Fig. 2. Partial training samples. (a) With four defects, one gap, one scratch, and two speckles; (b) with one speckle defect; (c) with two gap defects

图 3. 部分测试样本。(a)包含一个缺口缺陷;(b)包含两个缺陷,即缺口和斑点;(c)包含一个划痕缺陷

Fig. 3. Partial test samples. (a) With one gap defect; (b) with two defects, one gap and one speckle; (c) with one scratch defect

图 4. 基于ZF网络的训练损失。(a) RPN第一阶段训练损失;(b) Faster RCNN第一阶段训练损失;(c) RPN第二阶段训练损失;(d) Faster RCNN第二阶段训练损失

Fig. 4. Training loss based on ZF network. (a) Stage-1 training loss of RPN; (b) stage-1 training loss of Faster RCNN; (c) stage-2 training loss of RPN; (d) stage-2 training loss of Faster RCNN

图 5. 改进后ZF网络的训练损失。(a) RPN第一阶段训练损失;(b) Faster RCNN第一阶段训练损失;(c) RPN第二阶段训练损失;(d) Faster RCNN第二阶段训练损失

Fig. 5. Training loss based on improved ZF network. (a) Stage-1 training loss of RPN; (b) stage-1 training loss of Faster RCNN; (c) stage-2 training loss of RPN; (d) stage-2 training loss of Faster RCNN

图 6. 使用labelImg标注数据集

Fig. 6. Labeling data set with labelImg

图 7. 马克杯缺陷检测结果对比图。(a)原始Faster RCNN检测结果;(b)加入FPN后的Faster RCNN检测结果

Fig. 7. Comparison of mug defect inspection results. (a) Original Faster RCNN; (b) Faster RCNN after FPN addition

表 1改进前后的ZF网络结构

Table1. ZF network structure before and after improvement

表 2不同网络结构的分类性能

Table2. Comparison of various network structures on classification performance

关于本站 Cookie 的使用提示

全站搜索

基于改进Faster RCNN的马克杯缺陷检测方法 下载： 1395次

图 1. FPN+RPN算法网络结构图

Fig. 1. FPN+RPN algorithm network structure diagram

图 2. 部分训练样本。(a)包含四个缺陷,即缺口、划痕、两个斑点;(b)包含一种缺陷,即斑点;(c)包含两个缺口缺陷

Fig. 2. Partial training samples. (a) With four defects, one gap, one scratch, and two speckles; (b) with one speckle defect; (c) with two gap defects

图 3. 部分测试样本。(a)包含一个缺口缺陷;(b)包含两个缺陷,即缺口和斑点;(c)包含一个划痕缺陷

Fig. 3. Partial test samples. (a) With one gap defect; (b) with two defects, one gap and one speckle; (c) with one scratch defect

图 4. 基于ZF网络的训练损失。(a) RPN第一阶段训练损失;(b) Faster RCNN第一阶段训练损失;(c) RPN第二阶段训练损失;(d) Faster RCNN第二阶段训练损失

Fig. 4. Training loss based on ZF network. (a) Stage-1 training loss of RPN; (b) stage-1 training loss of Faster RCNN; (c) stage-2 training loss of RPN; (d) stage-2 training loss of Faster RCNN

图 5. 改进后ZF网络的训练损失。(a) RPN第一阶段训练损失;(b) Faster RCNN第一阶段训练损失;(c) RPN第二阶段训练损失;(d) Faster RCNN第二阶段训练损失

Fig. 5. Training loss based on improved ZF network. (a) Stage-1 training loss of RPN; (b) stage-1 training loss of Faster RCNN; (c) stage-2 training loss of RPN; (d) stage-2 training loss of Faster RCNN

图 6. 使用labelImg标注数据集

Fig. 6. Labeling data set with labelImg

图 7. 马克杯缺陷检测结果对比图。(a)原始Faster RCNN检测结果;(b)加入FPN后的Faster RCNN检测结果

Fig. 7. Comparison of mug defect inspection results. (a) Original Faster RCNN; (b) Faster RCNN after FPN addition

表 1改进前后的ZF网络结构

Table1. ZF network structure before and after improvement

表 2不同网络结构的分类性能

Table2. Comparison of various network structures on classification performance

相关论文

相关资讯

关于本站 Cookie 的使用提示

全站搜索

基于改进Faster RCNN的马克杯缺陷检测方法下载： 1395次