4results about How to "Optimize the total calculation amount" patented technology

The invention relates to the technical field of medical image processing, and discloses a large kernel decomposition attention network for lung CT image super-resolution reconstruction. The invention provides a novel multi-scale large-kernel decomposition attention (MLDA), and the detail features of various structural tissues in a lung CT image are captured by combining a multi-scale strategy and the large-kernel decomposition attention (LDA). Meanwhile, the LDA avoids block artifacts caused by using expansion convolution in large kernel decomposition. In addition, the MLDA provides three receptive fields with different sizes, so that the MLDA can flexibly cope with multi-granularity feature extraction from micro tissues to large lung anatomical structures. According to the method, the MLDA serves as the Token Mixer to be integrated into the MetaFormer architecture, local lung anatomical structure details can be captured, and the recovery quality of the whole image can be improved. Compared with a traditional convolutional neural network, the combination of the MetaFormer architecture and the MLDA enables the model to model local features and a remote dependency relationship at the same time, and maintains a good calculation amount.

The invention belongs to the technical field of radar signal processing, and particularly relates to a radar communication integrated clutter suppression method, a radar communication integrated clutter suppression system, radar communication integrated clutter suppression equipment and a medium. And a radar communication integrated echo signal model containing clutter, target echo and noise is designed based on the model. The radar communication integrated signal waveform with the low sidelobe characteristic is obtained through iterative solution by designing an optimization problem taking an integral sidelobe ratio matched with a filtering result as a cost function. Meanwhile, a two-stage cascade filter is designed to filter echo signals. And finally, performing communication information compensation on the filtered echo signal, and performing target parameter estimation by adopting a two-dimensional FFT algorithm and a subspace algorithm. The method effectively reduces sidelobe interference, improves the precision of target detection, and is suitable for high-resolution radars and high-speed communication systems.

The application provides a dynamic equalization method and circuit, which comprises the following steps: calculating a maximum gain coefficient and a minimum gain coefficient according to an input signal and preset processing parameters; multiplying the filtered input signal and one of the maximum gain coefficient and the minimum gain coefficient to obtain a first output signal; multiplying the input signal directly or after filtering or delaying and the other of the maximum gain coefficient and the minimum gain coefficient to obtain a second output signal; and superimposing the first output signal and the second output signal to obtain a final output signal; and adjusting the proportional relationship between the two input signals to achieve the purpose of adjusting the final output signal in real time. This method simplifies the required calculation amount and storage space of the dynamic filter, avoids the problem of harmonic and noise caused by the sudden change of the filter coefficient, and is effective in optimizing the calculation amount.

The invention discloses an online domain adaptive learning method based on multi-head self-integration, and relates to the field of machine learning. The method comprises: pre-training a backbone network of a multi-head model; respectively training each classification head of the multi-head model; performing TTA of multi-head model self-integration on the noise data set; performing optimal integration weight search by using Bayesian optimization; and updating and optimizing the weight of the multi-head model. According to the method, the precision of the model on a distributed out-of-data set is effectively improved, and the practical application capability of the model is enhanced under the condition that the parameter quantity and the calculation quantity of the model are not remarkably increased; and an efficient and practical solution is provided for online domain adaptive learning.