summaryrefslogtreecommitdiff
path: root/macros/latex/contrib/hitszthesis/tex/abstracten.tex
diff options
context:
space:
mode:
Diffstat (limited to 'macros/latex/contrib/hitszthesis/tex/abstracten.tex')
-rw-r--r--macros/latex/contrib/hitszthesis/tex/abstracten.tex18
1 files changed, 18 insertions, 0 deletions
diff --git a/macros/latex/contrib/hitszthesis/tex/abstracten.tex b/macros/latex/contrib/hitszthesis/tex/abstracten.tex
new file mode 100644
index 0000000000..bfcc60d8ca
--- /dev/null
+++ b/macros/latex/contrib/hitszthesis/tex/abstracten.tex
@@ -0,0 +1,18 @@
+% !TEX root = ../main.tex
+
+% 英文摘要
+\begin{abstracten}
+
+In recent years, the research of intelligent object grabbing by robots based on neural network method is in the ascendant. Firstly, this paper introduces the background of this research, and it is a feasible scheme to apply the neural network method to the intelligent grasping of robots.
+
+Then, this paper summarizes the research history and status quo of neural network in the field of robotic intelligent grasping. From the three milestone models of real-valued neural networks to the emergence of complex-valued neural networks, the model of neural networks has been greatly expanded. The method based on deep learning has made great achievements in the field of robot vision recognition. After recognizing the object image to be grasped, the strategy of robot grasping is also very exquisite.
+
+This paper also focuses on the intelligent grasping of robots based on three different neural network methods. Firstly, the method based on complex-valued neural network includes three models: discrete, continuous and multi-valued. Complex-valued neural network, with its natural complex processing ability, makes the signal which often needs to be processed in frequency domain, such as image, have a direct expression and processing mode.
+
+Then it is based on the method of lightweight convolution neural network. At present, the network structure used in the research of robot grasping pose prediction method based on neural network usually has a large number of parameters, which requires a lot of computing and storage resources. The grabbing prediction model of lightweight convolutional neural network based on SqueezeNet has smaller network model, less storage resources and faster speed without reducing accuracy.
+
+Finally, based on the method of cascade convolution neural network, a convolution neural network model Angle-Net for fine grabbing attitude estimation is proposed. On this basis, a two-stage cascade grabbing pose detection model is proposed.
+
+% 英文关键词
+\keywordsen{Lightweight, Convolutional Neural Network, Complex Neural Network, Two stage cascade grabbing pose detection model}
+\end{abstracten} \ No newline at end of file