基于机器视觉的八角颜色与果形识别研究

为了提高八角外观检测精度及效率,基于机器视觉技术对不同果形和颜色的八角进行识别和处理。果形在RGB、HSI颜色空间中,根据棕红、黑红、褐红在H颜色空间的区别,提取了不同颜色的八角H分量值,识别正确率为95.12%、95.12%、97.56%。利用极坐标变换思想建立极坐标模型,通过对极坐标模型错位相减、归一化、角数判别,有效的识别八角的角数,识别正确率为94.73%;在角数识别的基础上通过余弦定理实现了粗短八角角瓣、瘦长八角角瓣的判别,识别正确率分别为94.29%、97.14%;通过对极坐标变换后的轮廓进行傅里叶变换识别了粗短八角角瓣和瘦长八角角瓣,识别正确率分别为94.29%、94.29%;通过对八角波峰点进行标准差分析,有效地识别了八角是否均匀。以上方法识别率高且精确,为八角的外观检测技术提供了理论基础和前景。

Research on Octagon Color and Fruit Shape Recognition Based on Machine Vision

In order to improve the octagonal appearance detection accuracy and efficiency, machine vision technology was used to recognize and process different fruit shapes and colors of star anise. In RGB and HSI color space, according to the difference of brown-red, black-red and Maroon in H color space, the H component values of different colors of star anise were extracted, and the recognition ratios were 95.12%, 95.12% and 97.56%, respectively. The polar coordinate model was established by using the idea of polar coordinate transformation. By dislocations subtraction, normalization and angle number discrimination of the polar coordinate model, the angle number of the octagon was effectively identified, and the recognition proportion was 94.73%. On the basis of angle number recognition, the discrimination of short and thick octagonal lobe and long and thin octagonal lobe was realized by the law of cosine, and the recognition proportion was 94.29% and 97.14%, respectively. The short and thick octagonal lobe and long and thin octagonal lobe were identified by Fourier transform of the contour after polar coordinate transformation, and the recognition ratio was 94.29% and 94.29%, respectively. Through the standard deviation analysis of the peak point of anise wave, the uniformity of anise wave was effectively identified. The recognition rate of the above method is high and accurate, which provided a theoretical basis and prospect for the appearance detection technology of star anise.