High Precision Calibration Algorithm for Binocular Stereo Vision Camera using Deep Reinforcement Learning

A machine vision-based wear detection method for guide pair of the mine hoisting system

The wear detection of the guide pair (GP) plays a key role in the safe operation of the mine hoist system. Due to the actual working conditions of the well, manual detection is still the main detection method for GP wear, which has the problems of time consumption, low detection accuracy, and being unable to realize real-time detection. In view of this situation, this paper studies a machine vision-based wear detection method of GP in a mine hoisting system. First, the wear detection algorithm of GP is designed by means of image correction, image preprocessing, and edge extraction. Then, the hardware of the detection system is selected and designed, and the interface of the upper computer is designed by LABVIEW. Finally, according to the actual underground working conditions, a test platform for the wear detection system is built, and the detection experiment is carried out. The results show that the method can detect the wear and the location of the GP's wear in real time. The maximum average error of the detection under three different wear conditions is 3.54%, which meets the requirements of the specified measurement accuracy. It can provide technical support for the automatic detection of the wear of GP in mine hoisting systems.

Large gear structure assembly method based on uncalibrated image visual servo guidance

The reliability of the transmission system hinges significantly on the assembly quality of its main component, the large gear structures. However, the traditional approach of employing manual lifting presents a host of challenges, such as high assembly complexity and lowered efficiency, rendering the overall assembly process notably arduous. In this study, a large gear structure assembly method based on uncalibrated image visual servo guidance is proposed. Comprising three modules, the approach involves constructing a task function for projective homography, estimating the image Jacobian matrix, and designing an adaptive servo controller. This methodology facilitates the mapping of changes in gear images to the motion of the end-effector in the parallel mechanism. Consequently, the system dynamically guides the end-effector to achieve the required attitude adjustments in the gear assembly in response to changes in the image features. Experimental results demonstrate that the method proposed surpasses alternative approaches, simultaneously exhibiting a significant enhancement in assembly efficiency. The method has a wide application prospect in the field of automated assembly of large gear structures.

Research on calibration of a binocular stereo-vision imaging system based on the artificial neural network

Camera calibration is a key problem for 3D reconstruction in computer vision. Existing calibration methods, such as traditional, active, and self-calibration, all need to solve the internal and external parameters of the imaging system to clarify the image-object mapping relationship. The artificial neural network, which is based on connectionist architecture, provides a novel idea for the calibration of nonlinear mapping vision systems. It can learn the image-object mapping relationship from some sample points without considering too many uncertain factors in the middle. This paper discusses the learning ability. A binocular stereo-vision mapping model is used as the learning model to explore the ability of image-object mapping for artificial neural networks. This paper constructs sample libraries by pixel and world coordinates of checkerboard corners, builds the artificial neural network, and, through the training samples and test samples prediction, verifies the learning performance of the network. Furthermore, by the laser scanning binocular vision device constructed in the authors' laboratory and trained-well network, the 3D point cloud reconstruction of a physical target is performed. The experimental results show that the artificial neural network can learn the image-object mapping relationship well and more effectively avoid the impact of lens distortion and achieve more accurate nonlinear mapping at the edge of the image. When the X and Y coordinates are in the range of 100 mm and the Z coordinates are in the range of a 1000 mm, the absolute error rarely exceeds 2.5 mm, and the relative error is in the level of 10^-3; for 1000 mm distance measurement, the standard deviation does not exceed 1.5 mm. Network parameter selection experiments show that, for image-object mapping, a three-layer network and increasing the number of hidden layer's nodes can improve the training time more significantly.