面向珍珠分拣机器人的形状视觉检测方法

Shape Visual Inspection Method For Pearl Sorting Robot

摘要:

针对人工进行珍珠形状分拣效率低、精度不稳定等问题，提出基于机器视觉的珍珠形状检测方法。采用背光成像方式消除珍珠表面纹理和光泽的影响，对获取的珍珠图像进行同态滤波等预处理算法，提高图像对比度。为了解决相互接触珍珠影响珍珠轮廓提取的问题，采用分水岭算法对珍珠图像进行分割，得到了独立存在的珍珠个体，再通过连通域标记、质心算法对珍珠进行定位。根据国家标准对珍珠形状的规定，基于珍珠图像信息建立珍珠形状参数模型，对珍珠形状进行量化。实验结果表明，不同形状的珍珠样本的检测误差为0.63%，形状统计精度为100%，算法耗时24 ms。该方法可准确高效对珍珠进行分拣分级，具有一定的实用价值。

Abstract:

Aiming at the problems of low efficiency and unstable precision of manual pearl shape sorting, a pearl shape detection method based on machine vision is proposed. The backlight imaging method is adopted to eliminate the influence of pearl surface texture and luster, and pre-processing algorithms such as homomorphic filtering are performed on the acquired pearl image to improve the image contrast. In order to solve the problem that the contacting pearls affect the extraction of the pearl contour, the watershed algorithm is used to segment the pearl image, and the independent pearl individual is obtained, and then the pearl is located by the connected domain mark and the centroid algorithm. According to the national standards on pearl shape, the pearl shape parameter model is established based on the pearl image information to quantify the pearl shape. Experimental results show that the detection error of pearl samples of different shapes is 0.63%, the shape statistics accuracy is 100%, and the algorithm takes 24 ms. This method can sort and classify pearls accurately and efficiently, and has certain practical value.