It is an important solution to construct a lookout support picture recognition neural network system for navigation at sea in order to raise a ship operation support level. This paper is concerned with the recognition map development for a target ship's course estimation and for a target ship's situation estimation in nighttime. The neural network input data is navigation lights location's binary address which is calculated a result of lapped navigation environment picture over the recognition map. The authors have constructed four kind of recognition maps which have different concepts one another. Characteristics of the neural network systems which have the different recognition maps are analyzed and four guidelines of the recognition map building are given.