Abstract: This paper has advanced an improved algorithm based on mathematical morphology and histogram projection.In the locating section，an improved algorithm based on edge detection and mathematical morphology is used to connect the container code area into a whole connecting region，which can fix the size of the morphologic structure adaptively and solve the problem of vertical array of the container characters.In the segmentation section，mathematical morphology is employed to remove noise，and histogram projection is used to segment each character horizontally and vertically.The experiment results indicate that this algorithm is easy and available with little prior information and the overall segmentation accuracy can achieve 93.33%，which proves the validity of the algorithm.

Key words: edge detection, mathematical morphology, connected region, histogram projection

摘要： 提出了基于数学形态学和直方图投影的集装箱箱号分割改进算法。在箱号定位阶段，运用基于边缘检测和数学形态学的改进算法，能自适应确定形态学结构元素的大小，将箱号区域连通成一个区域，并能解决集装箱文字纵向排列的问题。字符分割阶段用数学形态学方法消除干扰边缘和噪声，通过投影直方图法完成行与列分割。实验表明，该算法简单可行，只需较少的集装箱先验信息，并且整箱分割正确率达到93.33%，证明了算法的有效性。

关键词: 边缘检测, 数学形态学, 连通区域, 直方图投影