Abstract: Rice printing planter is one of the effective methods for precise and uniform sowing of rice seeds. Due to the influence of mechanical vibration, seeding paper speed, and rice seed transportation, several rice seeds would adhere to some glue points on the seeding paper, which affected the seeding quality. To address this issue, a seed rate detection method based on machine vision for printing planters was proposed in this paper. Firstly, the image information of the seeding paper was collected by a CCD camera. Then, the redgreen color difference was used to extract the seed color information, morphological processing methods such as Otsu threshold segmentation were used to separate the seeds from the background, and an opening operation was utilized to separate slightly sticky rice seeds. Additionally, the area threshold method was adopted to remove noise and accurately determine the number of rice seeds. Finally, the ratio between the total area of tightly adhered rice seeds and the average area of each rice seed was used to correct the number of rice seeds. The results show that the seed rate detection method is effective in obtaining the number of rice seeds from the glue point of the seeding paper. The range of seeds rate detected was 88.84%-92.05%.

Key words: rice, printing planter, machine vision, adaptive morphology, precise positioning seeding, seeds rate

摘要： 水稻印刷播种是实现稻种精确定位和均匀播种的有效方法之一。由于机械振动、播种纸走速、稻种输送等环节影响，播种纸的某些胶点上可能会粘附多个水稻种子，影响播种质量。为此，提出基于机器视觉的水稻印刷播种机成种率检测方法。首先，利用工业CCD摄像机采集播种纸正面图像信息；其次，利用红绿色差提取种子的颜色信息，采用Otsu阈值分割等形态学处理方法实现种子与背景的分离，并利用开运算实现轻微粘连稻种的分离；然后，利用面积阈值方式去除噪点，从而实现水稻种子数量的统计；最后，根据粘连稻种总面积与每个稻种平均面积的比值对紧密粘连的稻种进行处理，从而修正水稻种子的数量。研究结果表明，该成种率检测方法是有效的，能够从播种纸的胶点上获取水稻种子数量信息，检测的成种率范围介于88.84%~92.05%。

关键词: 水稻, 印刷播种机, 机器视觉, 自适应形态学, 精确定位播种, 成种率