DNA barcoding has become the most important technique in species identification of plant materials for its high repeatability and accuracy. However, the successful application of this technology is dependent on the reference library of high species coverage. Unfortunately, such a library still remains to be constructed due to shortages of reliable plant materials, funds and experimental and data processing skills. Following our previous suggestion of using herbarium materials for fragments of ca. 400bp, we sampled 380 flowering plants, sequenced six fragments of conventional DNA barcodes (ITS1, ITS2, matK1, matK2, rbcL1 and rbcL2) by two rounds of high throughput PCR with unique 10bp oligoes to label samples on NGS platforms (Illumina HiSeq 2500 and Ion Torrent S5) and Sanger sequencing platform (ABI 3730xl), and analyzed the NGS data with our new Cotu method together with other three methods. After comparing sequencing depth variations among samples, base qualities, read lengths, sequence accuracies and costs of the resulting sequences by NGS to the references sequences by Sanger sequencing, we think Illumina HiSeq 2500 PE250 run is more suitable for DNA barcoding. Based on the comparisons of recoveries, lengths, accuracies and reliability of resulting sequences created by Cotu and other three methods to the references, we recommend Cotu as the first choice for NGS data analyses for its longer resulting lengths, higher sequence accuracy and more ease to handle. With the herbarium materials being used, high throughput machines (PCR and NGS) being employed, and Cotu method being adopted, the money and labor investments will be significantly reduced and a regional or global DNA barcoding reference library with high species coverage is likely to be built up in a few years.