Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

< Back to Article

Fig 1.

The system architecture of the grinding station.

More »

Fig 1 Expand

Fig 2.

The layout of the grinding station and unloading robot with vision system.

More »

Fig 2 Expand

Fig 3.

The strategy of robot workstation.

More »

Fig 3 Expand

Fig 4.

(a) The object and the region used for detection. (b) The contour of the object is not clear in the image.

More »

Fig 4 Expand

Fig 5.

Standard yolov3 network structure.

More »

Fig 5 Expand

Fig 6.

Part of the object in the dataset (from left to right: The original image of the object, image after shading adjustment, the image with Gaussian noise, blurred image, and rotation image).

More »

Fig 6 Expand

Fig 7.

Den-yolov3 network structure.

More »

Fig 7 Expand

Fig 8.

Creating the model map.

The model map is created by the local feature of the object. After the local feature is obtained by the Den-yolov3 detector, the geometric method and Line-MOD method is used to generate the model map.

More »

Fig 8 Expand

Fig 9.

Multilevel matching method (the input image is downsampled to level 4, then M4 set is used for template matching.

This determines the M3 model to be used and matches again until the level1).

More »

Fig 9 Expand

Fig 10.

The mesh region method is used to segment the burr area of the casting mouth (M is the amount of feed, it calculate by manual test).

More »

Fig 10 Expand

Fig 11.

Picture showing the grinding system.

More »

Fig 11 Expand

Table 1.

Training dataset.

More »

Table 1 Expand

Table 2.

Initialization parameters of the two networks.

More »

Table 2 Expand

Table 3.

Test result of stand yolov3 and Den-yolov3.

More »

Table 3 Expand

Table 4.

Initialization parameters of the two networks.

More »

Table 4 Expand

Table 5.

Accuracy of stand yolov3 and Den-yolov3.

More »

Table 5 Expand

Fig 12.

The loss function of stand yolov3 and Den-yolov3.

More »

Fig 12 Expand

Fig 13.

The test result of Den-yolov3 and stand yolov3: (a) Den-yolov3 test result. (b) yolov3 test result.

More »

Fig 13 Expand

Fig 14.

The object in this study test result of Den-yolov3.

More »

Fig 14 Expand

Table 6.

GMLINE-2D calculate result.

More »

Table 6 Expand

Fig 15.

Results of Line2D method result and GMLINE-2D.

More »

Fig 15 Expand

Fig 16.

Line-MOD and GMLINE-2D matching result.

More »

Fig 16 Expand

Fig 17.

Fast ICP registration for a feature area (X and Y coordinates are in pixels).

(a) the point after GMLINE-2D, (b) first time iterative, (c) second time iterative, (d) last time iterative.

More »

Fig 17 Expand

Fig 18.

Picture showing the object before (left) and after (right) grinding (the burr on the workpiece surface is cleanly removed).

More »

Fig 18 Expand