Detection of weld groove edge based on multilayer convolution neural network

Yang, Guowei; Wang, Yizhong; Zhou, Nan

doi:10.1016/j.measurement.2021.110129

Cited by 21 publications

(5 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Moreover, the development of novel wavelets are always in high demand for technological applications. [ 45 , 46 , 47 , 48 , 49 , 50 , 51 , 52 , 53 ] Our designed special function can be combined with time depended parameter that makes them very useful for constructing a new form of the wavelets.…”

Section: Discussionmentioning

confidence: 99%

Image Enhancement via Special Functions and Its Application for Near Infrared Imaging

et al. 2023

View full text Add to dashboard Cite

Image enhancement is important given that it can be used to highlight the area of interest in the images. This article designs four filters via special function for realizing image enhancement. Firstly, a filter based on the exponential function is designed. When the value of the progression is even, the edge feature can be extracted. When the value of the progression is odd, sharp contrast can be obtained. Secondly, a filter is built using hyperbolic cosine and its inverse function, where a printmaking feature can be extracted. Thirdly, a filter is made via a hyperbolic secant function and its inverse. It can lead to the extraction of image edge. When the progression value is increasing, marginal effect can be found and the brightness is decreasing. Ripple morphology can be found. Fourthly, a filter is constructed through a hyperbolic sine function and its inverse, where marginal features can be extracted. Furthermore, these filters are useful for extracting the marginal features even when a high noise density of 0.9 is added to the original images. They are useful for highlighting the images acquired from near infrared imaging.

show abstract

Section: Discussionmentioning

confidence: 99%

Image Enhancement via Special Functions and Its Application for Near Infrared Imaging

et al. 2023

View full text Add to dashboard Cite

show abstract

“…Deep convolutional neural networks use large amounts of data for feature extraction and have good feature representation and self-learning capabilities. Yang et al [8] designed a lightweight multilayer convolutional neural network in order to detect weld edges disturbed by noise. The network is capable of extracting multilayer features, which improves the resolution of weld edge detection and has a strong anti-interference capability.…”

Section: Introductionmentioning

confidence: 99%

Weld seam detection and location method based on depth camera and improved YOLOv5

zhang

2023

International Conference on Optics and Machine Vision (ICOMV 2023)

View full text Add to dashboard Cite

For automatic identification and positioning of the weld seam by the robot, this paper proposed a weld seam recognition and localization method based on the RealSense depth camera and the improved YOLOv5. Firstly, the original YOLOv5 model is improved by inserting the Coordinate attention model and getting the center point of the weld in the pixel plane according to the prediction frame. The actual position of the weld seam is then calculated by combining the depth information acquired by the RealSense depth camera. The test results show that the mAP index of this training model improves from 82.3% to 90.8%, which is significantly better than the model before the improvement. The maximum error is 2.9mm when identifying and positioning the object at a distance of 300mm, and the error percentage is within 2% when identifying and positioning the object at a distance of 0.3m-2m. This method established the relationship between the weld detection object and the robot position, compensating for manually moving the weld tracking sensor to its working range. It is of reference significance for welding robot automation.

show abstract

“…However, this approach is complicated because it requires gathering many images, it is difficult to establish an image fusion standard, and it may change the image's texture. Deep learning techniques may be used to determine suitable polarizing angles [22]. However, model training also needs many image samples in various lighting situations.…”

Section: Introductionmentioning

confidence: 99%

Camera calibration approach using polarized light based on high-frequency component variance weighted entropy for pose measurement

Zhang

et al. 2023

Meas. Sci. Technol.

View full text Add to dashboard Cite

The camera’s intrinsic parameters calibration directly affects monocular pose measurement accuracy. To improve the calibration accuracy of the camera’s intrinsic parameters, the polarizing device is used to eliminate the high-brightness areas in the images collected for camera calibration. However, an unsuitable polarizing angle will reduce the image’s contrast. This causes the texture information of the image to be lost or makes the extracted feature change in the direction of grayscale value decline, ultimately affecting the feature extraction. Aiming at this issue, this paper proposed a camera calibration approach using polarized light based on high-frequency component variance weighted entropy. We determined the suitable polarizing angle according to the 2D homography matrix describing the mapping between the target plane and normalized imaging plane for each calibration location. Under the suitable polarizing angle, we determined the optimal exposure time according to the high-frequency component variance weighted entropy. The images collected under suitable polarizing angles and the optimal exposure time were used to finish the camera calibration. The suitable polarizing angles and optimal exposure time can ensure the image’s contrast and texture information. The experimental findings demonstrate that our approach can increase the camera calibration accuracy and thus pose measurement accuracy. In the -45° to +45° range, the average pose measurement error is less than 0.05° and lowered by 0.08° (62%) compared to that performed using the intrinsic parameters calibrated with the images collected under no polarizing device. In the 0 to 20 mm range, the average pose measurement error is less than 0.028 mm and lowered by 0.05 mm (64%) compared to that performed using the intrinsic parameters calibrated with the images collected under no polarizing device.

show abstract

Detection of weld groove edge based on multilayer convolution neural network

Cited by 21 publications

References 18 publications

Image Enhancement via Special Functions and Its Application for Near Infrared Imaging

Image Enhancement via Special Functions and Its Application for Near Infrared Imaging

Weld seam detection and location method based on depth camera and improved YOLOv5

Camera calibration approach using polarized light based on high-frequency component variance weighted entropy for pose measurement

Contact Info

Product

Resources

About