Abstract: The article taking watermelon seeds as the research object，has been successfully applied into the automatic identification machine of watermelon seeds.By extraction of characteristic values，like the length of orthogonal long and minor axis of the watermelon seeds，the proportion between the length and width，the area，the degree of connecting superficial pattern，the coverage degree of black region and the distribution difference of histogram gradation and screening the watermelon seeds，by compared with the standard information database and melon seeds separation-equipment，the machine finally divides and picks the qualified melon seeds.The system takes visual interface.It has the interactive function between man and machine，being more objective，more accurate and more convenient，compared with the hands’ works.The realization of real-time image gathering function enormously enhances the speed and the efficiency.

Key words: orthogonal long and short axis, degree of connecting superficial pattern, distribution difference of histogram gradation, standard information database, melon seeds separation-equipment

摘要： 以西瓜子为研究对象，设计并实现了一系列特征提取算法和软硬件分捡流程，成功应用于西瓜子自动识别机。系统提取西瓜子正交长短轴长度、面积、表面花纹连通度、黑色区域覆盖度、直方图灰度分布差异等特征值，通过标准信息库比对及硬件分捡设备，最终选出合格瓜子。系统采用可视化界面，具备人机交互功能，相比人工分捡具有客观性强、准确度高、快捷方便的特点，特别是图像实时采集功能的实现，极大提高了分捡的速度和效率。

关键词: 正交长短轴, 花纹连通度, 直方图灰度分布差异, 标准信息数据库, 分捡设备