In order to avoid the deficiencies of conventional high voltage circuit breaker mechanical properties detection methods, a new algorithm based on image block matching with diamond search strategy is presented in this paper. The motion of auxiliary mark on the pull rod or shaft is firstly recorded by a high-speed and high-definition digital camera when the circuit breaker is open or close. Then the motion trajectory is acquired through diamond image block matching method. The mechanical parameters, such as travel and open and close velocity, are calculated according to the travel-time curve of the circuit breaker. Finally, evaluation model is constructed taking mechanical parameters characteristic values as inputs of ELM. Comparing to the existing techniques, our method is a noncontact measurement based on computer vision. It is easy and convenient for practical application since it need not any electrical and mechanical connection to the breaker. Another advantage of our method is that it can obtain the line and angle displacement simultaneously. The experiment results on the circuit breaker of 220 kv show that our method is effective for breaker mechanical properties detection.
Generally, a few-shot distribution shift will lead to a poor generalization. Furthermore, while the number of instances of each class in the real world may significantly different, the existing few-shot classification methods are based on the assumption that the number of samples in each class is equal, which causes the trained classifier invalid. Moreover, through ResNet and WRN (Wide Residual Network) have achieved great success in the image processing field, the depth and width of CNNs constrain the conventional convolution layer performance. Thus, to overcome the above problems, the model of this paper proposes a novel few-shot classification model that uses learning balance variables to decide how much to learn from the imbalance dataset, which dynamically generates the convolution kernel based on each input. In our model, to extend the decision boundaries and enhance the class representations, this paper uses embedding propagation as a regularizer for manifold smoothing. Manifold smoothing can effectively solve the above problems of transductive learning. The interpolations between neural network features based on similarity graphs are used by embedding propagation. Experiments show that embedding propagation can produce a better embedding manifold and our model in standard few-shot datasets, such as miniImagenet, tieredImagenet, CUB has state-of-the-art results. It significantly outperforms the existing few-shot approaches, which consistently improves the accuracy of the models by about 11%.
Existing quadtree-based fractal algorithms and fractal algorithms based on horizontal vertical (HV) have the problems of long encoding time and low accuracy in the task of image retrieval. In this paper, an improved fast fractal image retrieval algorithm based on HV segmentation is proposed, which speeds up the coding time and improves the accuracy for real-time searching. In order to improve the coding efficiency, the proposed algorithm restricts R block segmentation to certain direction and location in the coding phase and uses the local codebook to find the optimal matching of the partitioned blocks. We also introduce a weighting equation calculating method of area intersection to the image matching. New weighting parameters with respect to the sizes of partitioning blocks are proposed to improve the accuracy of image retrieval. The constraint-based HV segmentation algorithm and the local codebook matching strategy are tested on the texture and Olivetti Research Laboratory (ORL) face datasets. The experimental results show that the proposed algorithm accelerates the speed of image encoding. When the recall ratio is 100%, the precision of our algorithm has improved significantly. The proposed algorithm based on HV segmentation outperforms traditional fractal search algorithms in terms of adaption adaptivity.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.