<p>Fiber length is an important indicator of fiber quality in the textile industry, which plays a decisive role in the stability of the spinning process and the performance of the final product. Aiming at the problems of manual intervention, low efficiency and lack of accuracy in traditional length measurement methods, this paper proposes a fully automatic image measurement method based on a clustering algorithm with hemp fiber as the research object. The method firstly uses the skeletonization technology to pre-process the fiber image, and then through the clustering and cross-matching of fiber intersectons, it realizes the accurate identification and image reconstruction of multi-crossing, loop and dense crossing areas, so as to automatically measure the fiber length distribution and other key length parameters. The experimental results show that the errors of average fiber length, weighted length, upper quartile length, short fiber content (SFC) and uniformity index (UI) measured by this method are controlled within 1% ~ 2%, and at the same time, fibers that cannot be reliably identified due to severe entanglement are excluded from the length statistics, which significantly improves the measurement stability and reliability. This method not only greatly reduces the time and labor cost of manual measurement but also provides a new means of characterization that is efficient, accurate and easy to apply on a large scale for the quality inspection of hemp fibers.</p>