Based on the structure and dynamic tuning characteristics of the microwave cavity filter, the method of vision assisting mechanical arm motion is used for tuning. In order to meet the flexibility and high precision re...
详细信息
Based on the structure and dynamic tuning characteristics of the microwave cavity filter, the method of vision assisting mechanical arm motion is used for tuning. In order to meet the flexibility and high precision requirements of the microwave cavity filter tuning system, the thesis improves image recognition and hand-eye coordination methods. Image recognition using Otsu method for image segmentation, the relative 3 d position coordinates of the target screw are obtained by edge fitting circle and image sharpness, and the screw angle was extracted by dynamically obtaining the screw groove thread area, which was used to improve the applicability and accuracy of screw positioning and thread angle extraction. The hand-eye coordination part proposes a hand-eye coordination method based on image feedback. Through the feedback of the image, the PID control robot arm is continuously moved to make the screw center and the image center overlap to achieve hand-eye coordination. The experimental results show that compared with the traditional hand-eye calibration method, the method takes less time and has lower hand-eye coordination error, which greatly improves the flexibility and precision of the tuning system. Based on the improved method of image recognition and hand-eye coordination, it can realize accurate and fast tuning of microwave cavity filter, and it also has certain guiding significance for the tuning of similar industrial equipment.
This paper mainly studies the self-adaptation of Pulse Coupled Neural Network(PCNN) and the application in handwritten digit recognition. First, the edge extraction algorithm of image using PCNN and maximum entropy is...
详细信息
This paper mainly studies the self-adaptation of Pulse Coupled Neural Network(PCNN) and the application in handwritten digit recognition. First, the edge extraction algorithm of image using PCNN and maximum entropy is proposed, the parameters' optimization of PCNN is realized by Simple Genetic Algorithm. Then, the foveation algorithm based on PCNN is used to extract the feature points of handwritten digits. Finally, a BP neural network with two hidden layers is used to recognize the images of handwritten digits which have been preprocessed. Experimental results on handwritten digit recognition demonstrated that the proposed method reached good performance on feature extraction and the recognition has better accuracy compared with the original method using BP neural network directly.
The kernel learning algorithm has been widely used to solve the generalization problem existing in reinforcement learning. However, when the state-action space of the sample is large, the computation and storage burde...
详细信息
The kernel learning algorithm has been widely used to solve the generalization problem existing in reinforcement learning. However, when the state-action space of the sample is large, the computation and storage burden in the kernel learning algorithm will increase. In such case, there will be long running time and the real-time performance of the control system cannot be guaranteed. Therefore, in order to enhance the computational speed, this paper proposes a parallel architecture based iterative segmentation optimal cyclic block kernel learning algorithm. The experimental results show that the proposed method can significantly improve the computational efficiency of the kernel learning algorithm, which has important engineering practice significance.
In this paper, aiming at the pesticide residue in water resources, considering the necessity of rapid detection on site,a method based on machine vision for pesticide residue detection is proposed. The method combines...
详细信息
In this paper, aiming at the pesticide residue in water resources, considering the necessity of rapid detection on site,a method based on machine vision for pesticide residue detection is proposed. The method combines machine vision and image processing technology to quickly extract the RGB three-component eigenvalues of the pesticide residue detection card. Each RGB channel has 256 brightness levels, which can overcome the shortcomings of low brightness level and indistinct image,and improve the accuracy of detection. The results show that the method is simple,fast and efficient, it is suitable for quickly extracting the characteristic values of pesticide residue detection cards.
Video saliency detection aims to extract salient objects in video. In order to resolve the problems of incomplete extraction salient area and the mixing between background and salient objects at boundary, this paper p...
详细信息
Video saliency detection aims to extract salient objects in video. In order to resolve the problems of incomplete extraction salient area and the mixing between background and salient objects at boundary, this paper proposes a video saliency detection method via multi-features fusion based on Boolean Map. The main innovation in this method is that global topological relationship of video frames are utilized to compute saliency values. Firstly, boolean maps are generated according to the Boolean Map theory via combine motion features and color features in videos, and then attention maps are calculated by masking unsurroundedness area, finally, video saliency maps are obtained by fusing all attention maps. Experiments on Segtrack V2 and Fukuchi benchmark datesets show that the proposed method successfully obtains complete and clear boundary salient area, and it outperforms the other general models.
In this paper, the problem of synchronization is investigated for a class of discrete-time nonlinear singularly perturbed complex networks(SPCNs). A slow sampling discrete-time nonlinear SPCN model is first devised, w...
详细信息
In this paper, the problem of synchronization is investigated for a class of discrete-time nonlinear singularly perturbed complex networks(SPCNs). A slow sampling discrete-time nonlinear SPCN model is first devised, which includes the slow and the fast states as well as the general sector-like nonlinear functions. By utilizing the Kronecker product, a new Lyapunov function dependent on singular perturbation parameter(SPP) is constructed. A sufficient condition in terms of linear matrix inequalities(LMIs) is derived under which the discrete-time nonlinear SPCN is globally asymptotically synchronized. When these LMIs have feasible solutions, the global asymptotic synchronization is ensured and the upper bound of the SPP is evaluated. A numerical example is given to illustrate the effectiveness of our results.
Aiming at the problem that the target occlusion and target loss in the target tracking process can not be solved by most tracking algorithms, anti-occlusion correlation filter tracking based on multi-peaks response is...
详细信息
Aiming at the problem that the target occlusion and target loss in the target tracking process can not be solved by most tracking algorithms, anti-occlusion correlation filter tracking based on multi-peaks response is proposed. When the target is occluded, the response of the multi-peaks and the highest peaks exceeds the set thresholds, and the two-dimensional matrix of the Gaussian distribution is used to perform a point multiplication operation with the confidence map to obtain a new confidence map, and the current frame model update is stopped. Experiments show that compared with other algorithms, the algorithm has a significant improvement in tracking accuracy and speed when dealing with video sequences whose targets are occluded.
As a unique property of the object surface, the spectral reflectance plays an important role in computer vision applications and in realistic material reproduction. To determine the influence of the light source on th...
详细信息
As a unique property of the object surface, the spectral reflectance plays an important role in computer vision applications and in realistic material reproduction. To determine the influence of the light source on the spectral reflectance reconstruction accuracy, the pseudo-inverse method is adopted to reconstruct the spectral reflectance. Further, the genetic algorithm is used to optimize the light source for improving the reconstruction accuracy. The experiment results show that the spectral reflectance reconstruction accuracy is highly affected by the light source spectra, and the light source spectra optimized by the proposed strategy significantly outperforms several commonly used illumination source.
It is of great significance for accurate and fast magnetic measurement of high magneto-crystalline anisotropy film materials. In this paper, an surface magneto-optical Kerr effect automatic measurement system based on...
详细信息
It is of great significance for accurate and fast magnetic measurement of high magneto-crystalline anisotropy film materials. In this paper, an surface magneto-optical Kerr effect automatic measurement system based on LabVIEW is designed according to the practical requirements of hysteresis loop measurement of magnetic films. The system takes the SR830 lock-in amplifier as core, combined with virtual instrument technology and correlative detection technology, which achieves the effective measurement of the weak Kerr signal of Ni Fe soft magnetic film material under strong noise environment, and automatically draws its hysteresis loop. The measurement system has high resolution, high output signal-to-noise ratio, high automation, and has good application value.
Calligraphy is an important humanistic symbol of Chinese civilization. However, most of the calligraphy is incomplete, which has only a small number of Chinese characters circulating in the world. How to use these sam...
详细信息
Calligraphy is an important humanistic symbol of Chinese civilization. However, most of the calligraphy is incomplete, which has only a small number of Chinese characters circulating in the world. How to use these samples to efficiently restore the remaining calligraphy characters has always been considered as a difficult task. In this work, we propose Densenetpix2 pix model based style transfer method to solve this problem. By training some samples to learn the rules of transferring the printed font images to the calligraphy characters images, Densenet-pix2 pix can predict the remaining calligraphy characters. Our method modify the generation network and optimization strategies of style transfer, which improves the generation quality of the calligraphy characters and the stability of the model. In addition, we use pre-trained feature extraction models to extract content information and style information, and scientifically evaluate the quality of our generated calligraphy characters from these two aspects. We compared our method with several other baseline methods. The experimental results show that our method can effectively restore the remaining calligraphy characters, and the generated Chinese characters are more delicate.
暂无评论