Real time identification of extraneous noise in seismic surveys

A method of detecting and controlling noise in seismic surveys, comprising:

• receiving, by a data processing system comprising memory and one or more processors, responsive to a seismic wave generated by a source, seismic data detected by a sensor component of a seismic data acquisition unit;
• generating, by the data processing system, for a first plurality of windows of the seismic data, Hough tensors, wherein the Hough tensors are generated, for a first plurality of seismic data transforms in a plurality of dimensions;
• detecting, based on a comparison of an eigenvector and eigenvalue of a canonical matrix of the Hough tensors with a historical eigenvector and eigenvalue of a historical canonical matrix of historical Hough tensors of historical seismic data, a first presence of noise in the seismic data, wherein one or more eigenvalues of the canonical matrix comprise at least a first eigenvalues mapping the Hough tensors to noiseless historical data, and at least a second eigenvalues mapping the Hough tensors to abnormal historical data;
• obtain a machine learning model trained with historical seismic data comprising noiseless data and abnormal data;
• input one or more eigenvalues of the Hough tensors for new seismic data into the machine learning model to detect the first presence of noise in the seismic data;
• and
• providing, by the data processing system, responsive to detection of the first presence of noise in the seismic data, a notification to adjust a characteristic of the seismic survey to cause the new seismic data subsequent to the seismic data to have a second presence of noise that is less than the first presence of noise, wherein the notification is generated to delay or pause performance of the seismic survey, thereby improving the quality of subsequently collected seismic data and providing higher quality images.