基于生物特征纹理信息的西瓜追溯标识算法设计与实现

发布时间：2018-06-29 09:37

本文选题：生物特征 + 纹理　；参考：《农业工程学报》2017年18期

【摘要】：为了将生物特征识别技术应用于追溯系统中的果品标识,该文提出了一种采用西瓜蒂外围纹理信息标识果品个体的方法。该方法首先采集西瓜图像并在瓜蒂外围构造一个环形区域;将环形区域归一化后利用Gabor滤波器对图像纹理进行特征提取及编码;然后通过计算码间的Hamming距进行纹理编码的匹配。在试验中,为100个西瓜在采后贮藏初始、第7天和第14天各采集一幅图像并计算出3组每组100个纹理编码,将第7天、第14天的每一个编码与初始阶段的每一个编码两两比对,累计完成20 000次比对证实西瓜纹理具有唯一性,不同个体的纹理特征各不相同。使用成对数据的假设检验证实在采收后14 d内纹理特征不随时间的推移产生显著变化(P0.05)。通过理论和试验证实算法对平移、尺度和旋转具有适应性,即西瓜在图像中的平移、尺度和旋转对Hamming距的计算不会产生显著影响(P0.05)。使用最大类间方差法在50个西瓜的纹理编码内训练出判别阈值,并用该阈值判别另外50个西瓜3个时间点间相互比对生成的100个来自同一个西瓜的Hamming距和4 900个来自不同西瓜的Hamming距,结果表明上述Hamming距均能正确判别,准确率与召回率为100%。该研究为农产品追溯系统的标识技术提供新的思路。
[Abstract]:In order to apply biometric identification technology to fruit identification in traceability system, this paper proposes a method of identifying fruit individuals by using the texture information of watermelon pedicle periphery. Firstly, the watermelon image is collected and a ring region is constructed around Guarty. After normalization, Gabor filter is used to extract and encode the texture of the image, and then the matching of texture coding is carried out by calculating hamming distance between codes. In the experiment, 100 watermelons were stored at the beginning of postharvest storage. One image was collected on the 7th and 14th days, and three groups of 100 texture codes were calculated. Each encoding on the 7th and 14th day was compared with each other in the initial stage. The cumulative completion of 20,000 comparison shows that watermelon texture is unique, and the texture characteristics of different individuals are different. The hypothesis test using paired data confirmed that the texture features did not change significantly with the passage of time within 14 days after harvest (P0.05). Theoretical and experimental results show that the algorithm is adaptive to translation, scale and rotation, that is, the translation of watermelon in the image, scale and rotation have no significant impact on the calculation of hamming distance (P0.05). The maximum inter-class variance method was used to train the discriminant threshold in the texture coding of 50 watermelons. The hamming distance of 100 watermelons from the same watermelon and 4,900 hamming distances from different watermelons were determined by the threshold value. The results show that the hamming distances mentioned above can be correctly judged, and the accuracy and recall rate are 100. This research provides a new idea for the identification technology of agricultural product traceability system.
【作者单位】：华中农业大学信息学院;
【基金】：国家自然科学基金项目(31571351) 农业部农业技术试验示范项目(2015-2130106)
【分类号】：TP391.41

【相似文献】