37 results about "Denoising autoencoder" patented technology

The invention relates to a soil heavy metal element content predicting method based on a terahertz spectrum and a deep automatic encoder. Compared with the prior art, the method overcomes the shortcoming that ingredients of soil cannot be comprehensively analyzed and generalized on a large scale. The method comprises the following steps: acquiring and pretreating a soil sample; construction of a predicting model based on the deep automatic encoder: stacking a denoising automatic encoder and a compression encoder on construction of a model to form the predicting model of the deep automatic encoder; training of the predicting model of the deep automatic encoder: inputting terahertz spectrum data for training a sample into the predicting model of the deep automatic encoder to finish training of the predicting model of the deep automatic encoder; and predicting the heavy metal element content of soil. By a structural model of the deep automatic encoder and the terahertz spectrum, ingredients of the heavy metal elements of the soil are analyzed and predicted.

The invention discloses a fault isolation and identification method in a nonlinear process based on a denoising self-encoder. Use the sensor to collect training data and test data, construct statistics in the denoising autoencoder residual space and hidden feature space; calculate the process monitoring statistics of the test data, when the test data process monitoring statistics exceed the control limit of the process monitoring statistics, Then there is a fault; the objective function is established by introducing a regularization term, and the fault amplitude is obtained based on the adaptive moment estimation solution, and the fault amplitude of each sensor is used to judge the fault of each sensor, so as to realize the isolation and identification of the fault. The invention can meet the speed and accuracy requirements of large-scale process fault diagnosis in industrial process, and provides reliable and effective technical support for industrial production process control.

The invention discloses a wireless transmission method for high-dimensional damaged data based on a noise-reduction self-encoder. The method of the invention includes model training and end-to-end transmission. Model training firstly performs data preprocessing on the historical perception data set, and divides it based on the K-fold cross-validation method; then builds a noise reduction autoencoder model, and trains based on the proposed new noise adding method of introducing random Gaussian noise in batches Denoising autoencoder models. In end-to-end transmission, firstly, the trained noise reduction autoencoder is split into two parts and deployed at the sending end and the receiving end, and then preprocessing and dimensionality reduction operations are performed on the perception data of unknown types of noise interference at the sending end, and the dimensionality reduction The data is transmitted to the receiving end, and finally the reconstruction operation is performed on the receiving end to obtain the reconstructed data of the undamaged sensing data. The method of the invention can effectively perform dimension reduction transmission, noise reduction processing and reconstruction of high-dimensional damaged perception data, and can filter out noise interference and dimension reduction transmission when collecting data in harsh environments.

The invention discloses a SAR image change detection method based on a stacked semi-supervised self-adaptive denoising self-encoder, which solves the problem of low detection accuracy of coherent speckle noise points and many edge change regions in the existing method. The implementation steps are: first generate a multi-scale difference guidance map; use the phase 1 image as input to train SDAE; use the multi-scale difference guidance map, phase 1 and phase 2 images as input to train SSADAE, and SSADAE adaptive error function The weights obtained by SDAE training are used in the method; then SSADAE is used to calculate the feature vectors of phase 1 and phase 2 images; the two are subtracted to obtain the difference vector, and then FCM classification is performed on it to obtain the change detection result map. The present invention firstly proposes a multi-scale difference guidance map, which can highlight the change area in the difference map; the SSADAE proposed later can use a small number of marked samples in the image to further improve the accuracy of change detection.