TDLMS filter can be used for small target detection. However， the limitation of the structure of the filter and algorithm has created some difficulties in the detection. Here a TDLMS filter based on neighborhood analysis was proposed which would improve the structure of the input window and incorporate new data processing blocks. In this way more information would be obtained and used around the target， and more remarkable detection effects would be gotten. Experimental results prove the effectiveness of the proposed algorithm in detecting targets.