【摘要】 在工业领域中，工件边缘检测至关重要。但现有算法在这一领域表现不佳。为提高工件检测的准确率和效率，提出一种优化的Otsu-Canny模型用于工业环境下工件的边缘检测。首先，结合自适应滤波窗口和权重优化对高斯滤波方法进行了改进，以更好地去除噪声并保留边缘细节。其次，通过分析图像梯度直方图，采用分段采样和二分法改进了传统Otsu算法，并结合Canny算子以实现自适应最佳阈值选择。这一改进在保持结果准确性的同时显著降低了计算复杂性。在BSD500数据集上，该算法的F1分数显著高于其他算法并有较好的鲁棒性。对于两种不同工件，该方法所检测出的边缘轮廓较为完整且引入的干扰更少，实现正确定位的个数均高于其他算法。对于工件一，选择200个工件进行检测，准确放入模具个数为192。对于工件二，对500个工件进行检测，其中实现成功注油次数为483。这些结果有力地证明了该算法具有优越的鲁棒性、精确性和自适应性，满足实际工件检测需求。更多还原

【Abstract】 In the industrial field, edge detection of workpieces is crucial. However, existing algorithms perform poorly in this area. In order to improve the accuracy and efficiency of workpiece detection, an optimized Otsu-Canny model for edge detection of workpieces in industrial environments is proposed in this paper. Primarily, the Gaussian filtering method is improved by combining adaptive filtering windows and weight optimization to better eliminate noise and preserve edge details. Moreover, by analyzing the histogram of image gradients, the traditional Otsu algorithm is improved through segmented sampling and binary methods, and combined with the Canny operator to achieve adaptive optimal threshold selection. This improvement significantly reduces computational complexity while maintaining result accuracy. On the BSD500 dataset, our algorithm achieves significantly higher F1 scores compared to other algorithms and demonstrates good robustness. For two different types of workpieces, our method detects edge contours that are more complete with fewer introduced interferences. The number of successful precise localizations is also higher compared to other algorithms. For component one, 200 pieces were selected for inspection, and the accurate placement in the mold was 192. As for component two, out of 500 pieces inspected, the successful oil injection occurrences were 483.These results strongly indicate that our algorithm exhibits superior robustness, precision, and adaptability, meeting the requirements for practical workpiece detection.更多还原