In this paper, we present a new approach to recognize line segments based on run length connection. Four kinds of run length are defined to represent the edges. For each run length, its prior run length is searched to estabish run-length-chains according to the connectivity of the congener run length. The beginning and the end of the run-length-chains are stored in linked list,and the invalid nodes of the linked list are eliminated dynamically to save the memory space. After the following testing, the false candidates for line segments are eliminated,and the line segments and edge points satisfying the predefined rules are combined to ensure the validity and integrity of the extracted line segments. The method converts the complex line detection problem to a simple solution by constructing run- length-chains. The experiments on noise image show that our method has the merits of high real time,robustness to noise,small computational complexity and space complexity.