3results about How to "Simplify the algorithm process" patented technology

The application discloses a semi-supervised HRRP noise label filtering method based on early learning guidance, comprising the following steps: constructing an HRRP training data set and a deep neural network model; pre-training the deep neural network model; predicting the HRRP training data set by using the pre-trained network model and dividing the HRRP training data set into two data sets according to the confidence; constructing a known label data set and an unknown label data set to form a semi-supervised training data set and performing data enhancement; semi-supervised training the network model by using the enhanced data; generating the label of the unknown label data set by using the semi-supervised trained network model and fully supervised training the network model; and target recognition of the HRRP data by using the fully supervised trained network model. The application uses a generalized cross-entropy loss function in the early learning training stage of the model, proposes a data enhancement scheme suitable for HRRP, can better reduce the influence of noise labels on the model, and has better stability and implementability.

The invention provides an endoscope-based monocular high-precision quantitative defect detection device, and relates to the technical field of industrial detection, the endoscope-based monocular high-precision quantitative defect detection device comprises an endoscope, the endoscope is provided with a first hard shell, a second hard shell and a soft shell, the surface of the first hard shell is provided with an automatic calibration detection mechanism, the automatic calibration detection mechanism comprises a connector, and the connector is connected with the endoscope. The automatic calibration detection mechanism further comprises an interface position, a controller, a monitor, a computer, an optical fiber, a signal transmission cable, an image sensor and an objective lens, an endoscope is put into a detected object to take a picture, a picture matrix is segmented, a definition evaluation function is evaluated for a small picture target area in the center, whether the area reaches the optimal definition or not is judged, and if yes, the object is detected. And if the optimal definition evaluation result is reached, judging that the object distance of the measured target is available data, namely a picture with the optimal definition, and if the optimal definition evaluation result is not reached, continuing to move the lens of the endoscope until the optimal definition evaluation result is found so as to find the clearest picture.

The invention discloses an EMD-GRU-based long-period new energy power-risk prediction method and system, and the method comprises the steps: carrying out the standardization processing of an obtained new energy output power historical data set, carrying out the EMD decomposition of the standardized historical data set, obtaining different IMF sequences, analyzing the characteristics of IMF subsequences, and carrying out the prediction of the risk of the new energy output power. Then, taking the decomposed IMF power sequence as a flow index in an electric pinch point, taking a risk degree as an index for describing electric power quality in the electric pinch point, utilizing a GRU model to predict reserved subsequences respectively, and meanwhile, adding meteorological related factor data of a corresponding time point into an input variable; and outputting a prediction result corresponding to each sub-sequence, and superposing the prediction values of each sub-sequence to obtain a final result of new energy output power-risk prediction. According to the invention, the new energy consumption level can be improved, and the abandoned electric quantity can be reduced.