中文摘要：

平面、凹面、凸面是识别物体结构的重要线索,从单幅图像获取这些线索是进行三维重建、目标识别、场景分割和标定的关键步骤。针对这一问题,提出了一种基于灰度和轮廓特征的平面、凹面、凸面识别算法。通过区域的灰度拟合曲线分析区域的灰度分布特征与区域形状之间的关系,对区域轮廓进行方向编码,并将完整轮廓在角点处断开成若干连续线条,利用PCA方法判定其中是否包含光滑曲线段。综合考虑区域的灰度分布特征和区域轮廓特征,给出了区域表面形状的识别规则。灰度和轮廓两方面线索互相补充、互相约束,大大提高了区域表面形状的识别准确率。实验表明,该算法具有良好的识别效果。

英文摘要：

Flat, concave and convex surfaces play an important role in an object＇ s profile. Obtaining those surface features from a single image is a key step in three-dimensional reconstruction, target recognition, scene segmentation and calibration. To this end, the flat, concave and convex surface recognition algorithm based on gray-scale and contour features is proposed. Regional gray fitting curve is discussed. The regional gray distribution feature and the relationship between the regional profiles is determined. Regional contour directions are coded, the complete contours are terminated at angle points, then several continuous lines are formed. PCA methods are used to judge if smooth curves are contained. Regional gray distribution features and regional contour features are considered together and a recognition rule for the regional surface profile is proposed. It is found that the regional gray distribution and regional contour features complement each other, they also constraint one another. This relation significantly increases the recognition accuracy of the surface profile. This is shown through the agreement of both theoretical and experimental analysis.