Detection of weak targets in heavy ground clutter is one of key issues for Foreign Object Debris (FOD) surveillance radar on runways. Because of “Shelf-shielding” effect, traditional CFAR detection algorithms cannot effectively detect targets in multiple target environments. A method of trimmed-mean clutter-map CFAR detection is proposed. First, multiple targets echo model of Frequency Modulation Continue Wave (FMCW) radar based on the scenarios of airport runways and characteristics of radar system is established. Then, the largest samples in referencing cells are removed and mean of remaining samples is the clutter-level estimation. To obtain a stable detection threshold to detect targets, the clutter-map cells are updated scan-by-scan. At last, further analysis of main factors which affects detection performance of this algorithm is studied.