Copper tubular lug is a metal fitting used for power system connection, which is widely used in substations, power grids and other fields. It needs to have certain size specifications and appearance quality requirements to ensure safe and reliable electrical connection. However, since some appearance defects are inevitable during the production process, it is necessary to use automated machine vision technology for comprehensive inspection.
Difficulties in detection:
1. The material is copper, which will reflect light when imaging.
2. The color, texture, and finish of copper materials in different batches and parts are different. If the same detection standard is used for detection during the detection process, false detection will occur.
3. The defects on the metal surface are often subtle, requiring the image capture device to be highly sensitive.
4. The shapes of different metal accessories are often slightly different, which does not affect the use, but will be misjudged during machine vision detection.
How to solve these problems? Machine deep learning can solve them easily! Deep learning can learn from defective sample data and correct areas that are prone to misdetection, improving the accuracy and speed of subsequent product inspections. The following case study is from Intsoft Tech.
Detection scheme design:
Imaging system: Using high-resolution industrial cameras and multi-angle lighting design, the appearance details of the copper tubular lug can be fully captured. The relative position and angle of the camera and the object to be detected are optimized to minimize the impact of reflection and shadow on imaging quality.
Image processing: The original image collected is pre-processed by filtering, histogram equalization, etc. to improve image contrast and clarity, and prepare for subsequent defect identification.
Defect detection algorithm: Based on the deep learning model trained by Intsoft-AI, the classification and recognition algorithm for common defects of copper tubular lug is optimized, including surface depressions, scratches, oxidation spots, etc. And it compensates for the appearance differences of the inspected parts that do not affect the use, reducing false detection. The model has a high accuracy.
System integration: Integrate imaging, image processing and defect detection into an automated inspection system to achieve real-time monitoring of the appearance of copper tubular lugs, and automatically classify and sort them into different areas according to different appearance defects.
Detection process:
1. Original image:
2. Poor plating:
3. Surface scratch:
4. Severe deformation:
5. Shorter in length and width:
6. Without chamfer:
7. Dent:
8. Foreign matter in the hole:
