Traditional machine vision is relatively weak in the detection of defects such as scratches, oil stains and friction marks; The use of deep learning can effectively identify defects; In model training, it is often impossible to obtain a relatively stable model because of the small number of defect samples. At this time, adding an appropriate amount of positive samples and supporting the corresponding network can make the detection system maintain the stability of the system without reducing its sensitivity to defect types.
The full-size appearance detection system is a detection system designed for the battery module pack. The hardware part integrates a variety of visual sensors such as 2D plane array, 2D linear array, 3D linear array, etc; The algorithm adopts the detection technology combining traditional machine vision and deep learning; Finally, the test data will be effectively integrated with the MES system to achieve rapid deployment and convenient query of size and defect detection.

The full-size appearance detection system is a detection system designed for the battery module pack. The hardware part integrates a variety of visual sensors such as 2D plane array, 2D linear array, 3D linear array, etc; The algorithm adopts the detection technology combining traditional machine vision and deep learning; Finally, the test data will be effectively integrated with the MES system to achieve rapid deployment and convenient query of size and defect detection< p> Equipment advantages

Many types of inspection: 2D size, 3D flatness, scratches and stains

Strong data fusion: all test items are directly integrated with MES

Fast customization: the test content can be customized

Hardware composition

2D area array camera - information collection of scratches, oil stains, plane dimensions, etc

3D laser profiler - flatness and other height related information collection

2D linear array camera - profile, friction mark and other information collection