Plagiarism is a common problem faced by both academic and education fields. Although commercial plagiarism detection systems are relatively mature in terms of technology, they are not adopted in routine, real-time and lightweight fields such as student assignments detection because of high cost in efficiency and economy. We propose an extending classic Winnowing plagiarism detection algorithm, which can record the location and length while calculating the hash value of a text block. The location and length information in fingerprints can be used to locate and mark plagiarism text block in original documents. We describe algorithms for detecting, locating and plagiarism fingerprints index merging using the extended Winnowing, and performe some functional and performance experiments to test the algorithms. Experiments and actual running results show that the extended  Winnowing affects performance slightly, but it can meet the needs of small to medium applications under general hardware configuration. The extended Winnowing algorithm keeps the original features such as high efficiency, reliability and flexibility, and meanwhile gets improved in functionality and enhances its practicability and adaptability.