Abstract: Detecting buckwheat hulling is one of the key procedures to improve hulling efficiency and product quality during processing. Conventional detection of hulling performance can rely on manual sampling with visual inspection. However, manual approach cannot meet the efficient, standard and large-scale processing in buckwheat industry, due to the highly subjective, time-consuming and labor-intensive task. Therefore, it is of great practical significance to develop an efficient and reliable online detection for buckwheat hulling. In this study, the performance of an online detection device was evaluated for buckwheat hulling. A 6QB-150 rubbing-type buckwheat hulling machine was selected as the research object. Two modules were composed of: a sampling mechanism and an image acquisition. Among them, the sampling mechanism was used for timely and quantitative sampling of the hulled materials during operation, especially for the stable sample collection. The image acquisition was utilized to capture the high-resolution images of the sampled materials, thereby providing for the reliable visual data for subsequent analysis of hulling performance. The structural parameters of the sampling mechanism were optimized, such as the discharge height. Enhanced discrete element method (EDEM) was conducted to explore the motion behavior and spatial distribution of buckwheat particles during sampling. The simulation results demonstrated that the discharge height posed a major influence on particle dispersion. Once the discharge height was within the range of 550-600 mm, the buckwheat particles exhibited a uniform distribution, indicating the high consistency and accuracy of image acquisition. A series of experiments were conducted to further evaluate the performance and rationality of the online detection device under the different discharge heights. The experimental range was set from 350 to 600 mm with an interval of 50 mm. The distribution states of buckwheat samples were then captured at each height level using an industrial camera. The spatial distribution of the buckwheat grains during sampling was quantitatively evaluated to take the mean-to-variance ratio as an index. The experimental results indicated that discharge height shared the outstanding effect on the spatial distribution of particles and the quality of the captured images. The dispersion of buckwheat grains improved gradually, as the discharge height increased, leading to more uniform spatial distribution. Once the discharge height was controlled within the range of 550-600 mm, the mean-to-variance ratio of the captured buckwheat images ranged from 3.5 to 7.03, indicating low variability and high uniformity. Particle overlap and clustering were reduced to significantly enhance the clarity and reliability of image features for subsequent analysis. Overall, the stable and high-quality image data was achieved for the highly efficient online detection, compared with the conventional manual sampling. The relationship between discharge height and image acquisition quality can provide a theoretical basis and technical support to optimize sampling parameters of online detection. Furthermore, the research can contribute to the processing efficiency, product quality and equipment in modern agriculture.