目的 旨在传统Harris角点检测算子的基础上进行改进， 以提高算子的漏检率和伪角点检测能力。方法 以自动物流包装线上物料的识别为例， 把获取到的图像先进行预处理， 得到灰度图像， 在灰度图的基础上， 首先通过方向可调滤波器进行4个不同角度的旋转， 再分别进行角点检测， 最后通过逻辑运算综合判断真伪角点。结果 把图像预处理的图像数据使用改进后的Harris角点检测算子进行角点检测， 并与经典角点检测算子进行比较， 结果表明改进后的算子确实有很强的辨别真伪角点的能力。结论 实验证明该方法可有效提高角点检测算子的识别准确率， 误检率降低到了1.3%， 漏检率降低到了2.9%。

This work aimed to improve Harris operator based on the traditional Harris corner detection operator, in order to improve the detection accuracy and the false corner detection capability of the algorithm. Using the material recognition on automatic logistic packaging line as an example, the images obtained were first pretreated to get grey images, and rotation at four different angles was performed on the basis of the grey images using steerable filter. Then corner detection was performed, and finally, the authenticity of corner was comprehensively judged using logic computation. Corner detection of pretreated image data was conducted using the improved Harris corner detection operator, and compared to that using the traditional corner detection operator. The results showed that the improved operator indeed had very strong capability of judging the authenticity of corner. In conclusion, the experiment verified that this method could effectively enhance the recognition accuracy of corner detection operator, and the false detection rate in the experiment was reduced to 1.3% and the undetected rate was reduced to 2.9%.