Recognition algorithm of internal defect images of thermal battery

Table 1. Training parameter setting of BP neural network
View table
View in Article
Table 1. Training parameter setting of BP neural network
参数名称参数设置
输入节点 21
输出节点 5
训练算法 Levenberg-Marquardt算法
激励函数对称sigmoid函数
最大迭代次数 5000次
学习速率 0.05
收敛目标值 0.001

Table 2. Classification results of BP neural network
View table
View in Article
Table 2. Classification results of BP neural network
类别测试集数量/个准确率/% 总检测时间/s
正常 100 97 <0.1
漏装单体 100 98 <0.1
单体倒装 100 98 <0.1
次序错误 100 95 <0.1
漏装集流片 100 92 <0.1

Table 3. Classification results of CART decision tree
View table
View in Article
Table 3. Classification results of CART decision tree
类别测试集数量/个准确率/% 总检测时间/s
正常 100 96 <0.1
漏装单体 100 87 <0.1
单体倒装 100 90 <0.1
次序错误 100 92 <0.1
漏装集流片 100 94 <0.1

Table 4. Weight ratio of 2 kinds of classification algorithms
View table
View in Article
Table 4. Weight ratio of 2 kinds of classification algorithms
类别 BP神经网络分类权重 CART决策树分类权重
正常 0.51 0.49
漏装单体 0.53 0.47
次序错误 0.51 0.49
单体倒装 0.52 0.48
漏装集流片 0.49 0.51

Table 5. Experimental results of thermal battery detection accuracy

View table

View in Article

Table 5. Experimental results of thermal battery detection accuracy

实验编号	样本数	正确检测数	误判个数/个	漏检个数/个	检测准确率/%	平均检测时间/s
实验1	500	492	2	6	98.4	0.12
实验2	500	497	1	2	99.4	0.11
实验3	500	496	1	3	99.2	0.13
实验4	500	493	2	5	98.6	0.11
合计	2 000	1 978	6	16	98.9	0.12

Tools

Get Citation

Copy Citation Text

Wei ZHOU, Rui WANG, Fanqin MENG, Guoming JU, Qingyi MENG, Xu ZHANG. Recognition algorithm of internal defect images of thermal battery[J]. Journal of Applied Optics, 2022, 43(1): 60

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites