摘要： 为提高西瓜子自动筛选系统的性能，提出一种用于度量和处理西瓜子弯翘度的方法。把双目平行立体视觉与骨骼线相结合，利用双目立体视觉技术重建物体表面的三维信息。将该信息进行正交投影，分别得到西瓜子在前视图和侧视图的轮廓。利用曲率尺度空间与骨骼线相结合的方法对侧视图的瓜子轮廓进行特征提取。实验结果表明，该方法对弯翘瓜子有较高的识别精度。

关键词: 西瓜子, 立体标定, 立体校正, 三维信息, 骨骼线, 弯翘度, 曲率尺度空间

Abstract: To improve the watermelon seed automated screening system performance, this paper presents a method of watermelon seed combination parallel binocular stereo vision and bone line to use in measurement and process the warping degree of watermelon seed. This method uses binocular stereo vision technique for the reconstruction of three-dimensional object’s surface information. The three-dimensional information is projected in space rectangular coordinates, it can obtain the seed profile in the front view and side view. The combination of Curvature Scale Space(CSS) and bone line is used in feature extraction of the side view profile of the watermelon seed. Experimental results show that the method has high recognition accuracy for warped seed.

Key words: watermelon seed, stereo calibration, stereo rectification, three-dimensional information, bone line, warping degree, Curvature Scale Space(CSS)

中图分类号: 

• TP391