Adaptive detection problem of range-spread targets is investigated in generalized Pareto sea clutter. Texture component of non-Gaussian sea clutter is considered to follow inverse gamma distribution in order to match the non-Gaussian characteristic of clutter. In response to the heavy detection performance loss of adaptive detectors when the secondary data are limited, the persymmetric structure of the covariance matrix is merged into the proposed detector. The proposed adaptive detector can perform better in heavy spiky clutter. Meanwhile, experiments via exploiting simulated data and measured sea clutter data indicate that the proposed detector has superior detection performance, especially for limited secondary data.