ABSTRACT
The bearing ring has the characteristics of weak texture and high reflection, which easily cause errors in stereo vision matching. To address these issues, this paper proposes a two-stage spatial pose detection method for bearing rings based on YOLOv5 and binocular vision. In Stage 1 called Pre-positioning: the YOLOv5 is used to detect the unobstructed bearing ring; then binocular vision is employed to locate the unobstructed bearing ring; and a line laser projection is performed on the unobstructed bearing ring for fine positioning. In Stage 2 called Fine-positioning: the YOLOv5 is used again to locate the line laser stripes projected on the upper surface of the bearing ring, and then the detected line laser stripes are used to calculate the robot's best grasping pose. The practical application showed that the introduction of line laser solved the problem of weak texture matching, and the application of yolov5 solved the problems of high reflection on the bearing ring surfaces and strong noises after laser reflection. Experiments showed that YOLOv5 detected the unobstructed bearing rings and laser line stripes with 98.5% and 99.3% mAP, respectively, and this method had a 96.3% grasping success rate.
- Lenz, I., Lee, H., & Saxena, A. (2015). Deep learning for detecting robotic grasps. The International Journal of Robotics Research, 34 (4–5), 705–724. https://doi.org/10.1177/0278364914549607Google Scholar
- Bohg, J., Morales, A., Asfour, T., & Kragic, D. (2013). Data-driven grasp synthesis—a survey. IEEE Transactions on Robotics, 30 (2), 289-309.Google Scholar
- Morrison, D., Corke, P., & Leitner, J. (2020). Learning robust, real-time, reactive robotic grasping. The International journal of robotics research, 39 (2-3), 183-201.Google Scholar
- Collet, A., Berenson, D., Srinivasa, S. S., & Ferguson, D. (2009, May). Object recognition and full pose registration from a single image for robotic manipulation. In 2009 IEEE International Conference on Robotics and Automation (pp. 48-55). IEEE.Google ScholarDigital Library
- Wang, C., Cao, Y., Yu, F., & Tang, Y. (2021, February). Dynamic Weight of Adaptive Information Density Network for Image Super-Resolution. In 2021 2nd Asia Service Sciences and Software Engineering Conference (pp. 123-129).Google Scholar
- Xu, M., Wang, X., Zhangt, X., Bin, G., Jia, Z., & Chen, K. (2020, May). Computation-Efficient Multi-Model Deep Neural Network for Sleep Stage Classification. In Proceedings of the 2020 Asia Service Sciences and Software Engineering Conference (pp. 1-8).Google Scholar
- Lingyu, Z., Bin, W., & Bai, W. (2020, May). Call Attention to Stances: Detect Rumor with a Stance Attention Network. In Proceedings of the 2020 Asia Service Sciences and Software Engineering Conference (pp. 27-34).Google ScholarDigital Library
- Chen, W., Lei, Y., Liu, M., Zhang, J., & Yao, X. (2020, October). Workpiece Detection Based on Image Processing and Convolutional Neural Network. In 2020 International Conference on Sensing, Measurement & Data Analytics in the era of Artificial Intelligence (ICSMD) (pp. 302-306). IEEE.Google Scholar
- Lei, Y., Yao, X., Chen, W., Zhang, J., Mehnen, J., & Yang, E. (2020, July). Multiple object detection of workpieces based on fusion of deep learning and image processing. In 2020 International Joint Conference on Neural Networks (IJCNN) (pp. 1-7). IEEE.Google ScholarCross Ref
- Jiao, L., Zhang, F., Liu, F., Yang, S., Li, L., Feng, Z., & Qu, R. (2019). A survey of deep learning-based object detection. IEEE access, 7, 128837-128868.Google Scholar
- Zou, Z., Shi, Z., Guo, Y., & Ye, J. (2019). Object detection in 20 years: A survey. arXiv preprint arXiv:1905.05055.Google Scholar
- Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016). You only look once: Unified, real-time object detection. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 779-788).Google ScholarCross Ref
- Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012). Imagenet classification with deep convolutional neural networks. Advances in neural information processing systems, 25, 1097-1105.Google Scholar
- Simonyan, K., & Zisserman, A. (2014). Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556.Google Scholar
- Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., ... & Rabinovich, A. (2015). Going deeper with convolutions. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 1-9).Google ScholarCross Ref
- Ren, S., He, K., Girshick, R., & Sun, J. (2015). Faster r-cnn: Towards real-time object detection with region proposal networks. Advances in neural information processing systems, 28, 91-99.Google Scholar
- Redmon, J., & Farhadi, A. (2018). Yolov3: An incremental improvement. arXiv preprint arXiv:1804.02767.Google Scholar
Recommendations
An Efficient Pose Measurement Method for End Effector of Mobile Manipulator Based on Binocular Stereovision
2019 IEEE/ASME International Conference on Advanced Intelligent Mechatronics (AIM)The pose measurement for end effector of mobile manipulator in large-scale workspace is critical to processing quality control. In this paper, a novel low-cost measurement system without accumulative error and laser block is proposed. The measurement ...
Measurement of ice thickness based on binocular vision camera
2017 IEEE International Conference on Mechatronics and Automation (ICMA)This paper presents a new method to measure the ice thickness of the transmission line, which is based on the binocular visible light camera mounted on the UAV acquisiting the transmission line icing image, combined the infrared camera and word line ...
Binocular Vision Based Drivable Region Fast Detection for Indoor Mobile Robot
GCIS '13: Proceedings of the 2013 Fourth Global Congress on Intelligent SystemsA fast method based on binocular vision is proposed for mobile robot to detect drivable regions. First, the image is segmented into regions by searching contours. Second, part obstacle regions are determined by the vanishing line. Then, according to the ...
Comments