Ultra-low power consumption microcontroller trace point condensation method based on TensorFlow architecture

The invention discloses an ultra-low power consumption microcontroller plot condensation method based on a TensorFlow architecture, and the method comprises the steps: obtaining secondary radar original decoding data and calibration data of different maneuvering targets, and carrying out the correlation of the secondary radar original decoding data and the calibration data, and forming sample data; the method comprises the following steps: firstly filtering trace points in a target trace point group and amplitudes corresponding to the trace points through an RC low-pass filter, filtering out high-frequency noise brought by environment and hardware equipment, and shrinking the range of the trace point group; and inputting the trace point azimuth, distance and amplitude data into a three-dimensional k-means clustering model, and realizing accurate classification of a real trace point group and a trace point group formed by interference and calculation of the center of the trace point group through a k-means alg