摘要： Target detection in synthetic aperture radar image utilizing polarimetric information has attracted considerable attention. Single-target detector (STD), partial-target detector (PTD), and geometrical perturbation-polarimetric notch filter (GP-PNF) are three traditional polarimetric detectors based on polarimetric information. STD aims at detecting single targets, whereas PTD is suitable for partial targets. GP-PNF focuses on detecting targets with features, which are different from the homogeneous background. Both STD and PTD need a prior knowledge of the target, whereas GP-PNF needs to estimate the local clutter automatically. All these three methods use a feature vector to describe the character of the target or clutter. In fact, the feature vectors of the clutter and target may distribute in a subspace. Especially for the heterogeneous background, a feature vector cannot accurately describe the clutter. Motivated by this, this paper extends the clutter model from a complex feature vector to a complex feature subspace, which is suitable for a heterogeneously patched region and derives extended PTD and extended GP-PNF. Experimental results show the extended detectors’ validation and superiority to traditional detectors for target detection in heterogeneous regions.