Disclosed is a strip steel thickness lateral distribution characteristic parameter extraction method. The method is characterized by including: a1, describing a horizontal coordinate of original data, wherein the actually-measured thickness at an xi position is hi; a2, enabling p to be 3, q to be (n-2) and a minimum sum of square of deviations ymin to be 100; a3, taking xp and xq as boundary points; a4, listing fitting functions; a5, taking the situations that two-section function values are equal and first derivative values are equal as constraint conditions; a6, taking coefficients as variables, setting fitting curves and functions of sums of square of deviations and taking a minimum function value of the sums of square of deviations as a constraint condition; a7, solving the coefficients according to the simultaneous conditions and recording the sums of square of deviations of the functions and the actually-measured values; a8, enabling q to be (q-1), if q>=n/2, returning to the step a3, and if q<n/2, entering step a9; a9, enabling p to be (p+1), if p<=n/2, returning to the step a3, and if p>n/2, entering step a10; a10, outputting the coefficients as optimum strip steel thickness lateral distribution characteristic parameters. By the strip steel thickness lateral distribution characteristic parameter extraction method, steel strip edge drop is described, equipment provided with a profiler can be improved in the aspect of the fitting functions, and after equipment without the profiler performs lateral distribution extraction of the thickness offline, applying the method to online production enables strip steel sectional quality and fitting precision to be improved.