番茄不同病害类型的图像特征精准识别仿真

Simulation of Accurate Recognition of Image Features of Tomato Disease Types

番茄不同病害特征具有一定相似性,为准确获取病害类型,提出基于构造性核覆盖算法的番茄病害图像识别方法。使用均值滤波算法计算像素点加权平均值过滤噪声;经过开、闭运算,去除孤立点,改善边界模糊现象;采用最大类间方差法获取目标、背景的阈值,分割图像;利用RGB空间模型得到图像颜色特征,分别通过Hough变换与灰度级共生矩阵,获得图像边界与纹理细节,提取病害特征;选取输入图像集合与核函数,经过空间平面变换,将图像映射在核空间内确定覆盖中心,构建覆盖领域,完成图像识别。仿真结果表明,所提方法能够提高图像信噪比,针对不同病害类型,均有较高的识别精度。

Different tomato diseases have certain similar characteristics. In order to accurately obtain disease types, this article presented a method to recognize tomato disease images based on constructive kernel covering algorithm. At first, the mean filtering algorithm was used to calculate the weighted mean value of pixels and filter noise. After opening and closing operations, isolated points were removed, and then fuzzy boundaries were improved. Moreover, maximization of interclass variance was adopted to calculate the threshold of target and background, thus segmenting images. Furthermore, RGB spatial model was used to obtain color features of image. Respectively, Hough transform and gray co-occurrence matrix were used to obtain image boundaries and texture details and thus to extract disease features. After that, input image sets and kernel functions were selected, and the images were mapped into in kernel space through spatial plane transformation. Finally, the coverage center was determined, and the coverage field was constructed. Thus, we completed the image recognition. Simulation results prove that the proposed method can improve the signal-to-noise ratio of image and has high recognition rate for different diseases.