基于计算机视觉的玉米种子形态识别测量

Measurement of Maize Seed Morphological Recognition Based on Computer Vision

种子的一致性包括品质、形态和颜色特征方面。采用具有高度一致性的种子,可以提高后期的机械化作业程度,从而降低生产成本。要提高种子的一致性,则必须识别并清除裂变和畸形的种子,传统的清选操作只能保证种子的净度和健康度,且效率太低。计算机视觉是目前种子精选分级方面研究最多的技术,适用的作物种类也很广泛。为此,基于计算机视觉技术,建立了玉米种子的形态识别测量方法和系统。系统通过扫描获取种子图像,再进行预处理和图像分割,提取单粒种子的图像;然后以面积和圆形度为参数,将种子划分为4个等级。在验证试验中,系统的图像处理过程仅耗时2s,对种子大小和圆扁的识别准确率分别为98%和96%,能够满足自动分级的要求。系统的创新在于用扫描仪代替相机,分析的通量大,且成像的质量高。另外,仅选择两个特征值作为分级参数,减少了分析过程中的计算量,有利于在实时分级装置中应用。

The consistency of the seeds includes the quality,shape and color characteristics. The use of highly consistent seed,can improve the degree of mechanization of the latter part of the operation,thereby reducing production costs. To improve seed consistency,you must identify and remove fissures and deformed seeds. The traditional cleaning operation can only guarantee the clarity and health of the seed,and the efficiency is too low. Computer vision is currently the most studied in terms of seed selection grading,and the range of crops is wide. Corn is one of the three major food crops in China,and the selection of seeds is an important guarantee for high and stable yield of maize. Based on the computer vision technology,this paper established the method and system of morphological recognition of maize seed. The system obtains the seed image by scanning,preprocesses and image segmentation,and extracts the image of the single seed. And then the area and roundness as a parameter,the seeds are divided into four levels. In the verification test,the system image processing process takes only 2s,the seed size and round flat recognition accuracy rate of 98% and 96%,to meet the requirements of automatic classification. System innovation is to use the scanner instead of the camera,the analysis of the flux,and the high quality of imaging. In addition,only two eigenvalues are chosen as the classification parameters,which reduces the amount of computation in the analysis process and is beneficial to the application in the real-time classification device.