为实现自然条件下低分辨率苹果病害的智能识别,对获取图像进行预处理,采用改进的水平集交互式分割方法提取病斑.在实验的基础上,基于灰度共生矩阵提取8个纹理特征参数作为病斑的有效识别特征,构建了基于灰度关联分析的病害识别模型.实验结果表明,用优选的8个纹理特征和基于灰度关联分析识别模型,对3种病害的平均正确识别率最高达到85.41%,可以有效识别苹果的病害.%To realize intelligently identification of the apple fruit's disease using low- resolution image, after pre-processing, the diseased parts were extracted using improved level set interactive segmentation method. According to the experimental results, the co-occurrence matrix was used to extract eight texture characteristics as recognizable features. The disease identification model is constructed based on the gray relation analysis. Experiments show that the optimal 8 features and the gray relation analysis model have the average correct identification rate being 85. 41% for the 3 kinds of diseases, which can effectively identify the apple diseases.
展开▼