This paper processed the simple fractal data with pyramidal algorithm of wavelet analysis and, in combination with the "random resonance" theory, analyzed and approached noise and error in high-precision micro-magnetic anomaly. According to residual magnetic anomaly detected by tendency analysis, the authors suggest controlling normal random Brown motion of non-integer rank calculus with "shift with the window" least squares technique, and provide essential guidance for effective recognition of high-precision micro-magnetic anomaly.