【作者】 邓秀华；

【导师】 朱伟兴；

【作者基本信息】 江苏大学 ， 农业电气化与自动化， 2007， 硕士

【摘要】 随着除草剂广泛应用于除草中，虽然减轻了人工劳动量，但也带来成本增加、作物品质和农业生态环境的污染等问题。本文利用计算机图像处理技术和小波分析工具对小麦田间常见杂草进行边缘检测，为杂草识别的实现打下基础。针对田间摄取图像时的随机扰动，图像中含有的噪音，为了提高识别效果，本文对所获得的图像进行一些前期的预处理。首先要把用照相机获得的真彩色图像灰度化，然后用中值滤波法对灰度图像滤波，去除了高斯白噪声对图像的影响。为了使农作物杂草等绿色植物和土壤背景分离，采用阈值分割按灰度级分为若干部分，本文中为了减少计算量，将杂草图像变成仅含有目标和少量高频噪声的二值图像。本文重点就几种经典的图像边缘检测微分算子作了分析和优缺点对比，基于这些算子都是在原始图像上(时域)进行的，不能在频域上对信号分析，作者利用小波分析的“自适应性”和“数学显微镜性质”对图像检测边缘算法进行研究。第一种算法是基于B样条函数的多尺度分析和数学形态学理论，选择2阶B样条小波作为小波基函数，对杂草图像进行Mallat快速变换，为了避免滤掉弱的边缘，本文把B样条小波边缘检测算法和数学形态学边缘检测算法进行融合，最后得到了综合各个尺度特征的理想边缘。第二种算法是小波分析模极大值边缘检测方法，根据李氏指数与小波变换关系，采用小波模极大值在不同尺度下传播的特性，检测出图像在水平和垂直方向的极大值，然后利用模糊算法构造相应的隶属函数，提取弱边缘信息，最后得到不同尺度下的边缘图像，仿真结果表明这种方法可检测出弱边缘。第三种算法是基于小波包的边缘检测，小波包理论优于其它小波分析方法就在于它不仅对低频图像进行分解，对高频图像也是一样，这样就可以获得更多的图像信息，实验结果表明，经小波包分解后，重构得到的近似部分图像去除了高频分量，能够检测到原图像中检测不到的边缘。本论文对边缘检测算法进行了综合的研究、对比和改进，为可变除草剂喷洒提供了理论支持，有一定的实用价值。更多还原

【Abstract】 Herbicide as a kind of effective method widely used weed control, has reduced the amount of manual labor, however it also have brought a series of side effects such as the increasing production costs, crop quality problems and ecological environmental pollution issues. This paper tries to use the computer image processing technology and wavelet analysis to detect edge of the comman weeds in wheat fields. it will lay the foundation of the realization of weed identification.In view of the random perturbation such as containing a certain amount of noise in the image, in order to improve recognition results. Firstly, the true color image obtained by the camera should be preliminary pretreated into the gray. Then we use Median Filter to filt Gaussian white noise. To separate the crop plants, weeds and soil backgrounds, we divide the gray level by the threshold segmentation. To reduce the calculation amount of this paper, we will turn the weeds image into the binary image containing target and a small amount of high-frequency noise.In this paper, the advantages and disadvantages of several classic edge detection methods are firstly discussed and compared during the process of achieving Edge Extraction of weeds image. Wavelet transform is proposed to detect the image because of its "adaptability" and "mathematical microscope nature". The first algorithm is based on the number of B-spline function analysis and theory of mathematical morphology. Take the 2-order B-spline wavelet as wavelet mother function. To avoid removing the weaker edge during the process of Wavelet transform two-dimensional images of weeds based on the Mallat fast algorithm, this paper presents a kind of edge detection of integration of B-spline wavelet edge detection algorithms and mathematical morphology edge detection algorithm, then finally, get the ideal edge synthesized all standard features. The second algorithm is wavelet analysis modulus maxima edge detection method. In according to the relation of Lee index and wavelet transform, we utilize the transmission characteristics of wavelet maxima in different scales to detect out images maximum in the horizontal and vertical direction, then we construct the membership function with fuzzy algorithm to extract weak Edge Information, and finally we get the edge images of different scales. Simulation results show that this method can detect weak Edge. The third algorithm is based on wavelet packet edge detection. The wavelet packet theory is better than the other wavelet analysis. It is not only contained the low-frequency image decomposition but also contained the high-frequency images decomposition, through which way we can get more information. The results have showed that the reconstructed similar partial image can remove the high-frequency components and can be used to detect the edge which can not be detected on the verge of the original image after the wavelet packet is decomposed.In this paper we will make a comprehensive research, contrast and improvement of algorithm for edge detection, which will provide the theoretical support for variable herbicide spraying. The results in this research is of great practical value and of great importance to the agricultural development.更多还原