基于MATLAB的植物幼苗识别

论文总字数：17413字

目 录

摘要 1

Abstract 2

1绪论 3

1.1选题的目的和意义 3

1.2国内外研究状况 3

1.3本文的研究内容 4

2图像获取及图像分割 5

2.1图像的获取与数据库的建立 5

2.2 图像背景分离 5

2.2.1RGB颜色模型下的分量提取 5

2.2.2阈值分割法 7

2.2.3形态学处理 9

3图像特征的选择与提取 11

3.1图像特征的种类 11

3.1.1颜色特征 11

3.1.2纹理特征 11

3.1.3形状特征 13

3.2HOG特征算法分析与提取 13

3.2.1HOG特征原理 13

3.2.2颜色空间归一化 13

3.2.3梯度计算 14

3.2.4梯度方向直方图 14

3.2.5重叠块直方图归一化 15

4 分类器的选取与构建 17

4.1SVM基本原理 17

4.2SVM核函数 18

4.3分类系统设计 19

4.4实验结果 19

5结论和展望 20

5.1结论 20

5.2展望 20

参考文献 22

致谢 23

植物幼苗识别

应杰

，China

Abstract：There are many kinds of weeds in farmland, which seriously affect the production and yield of crops. Using image processing technology to distinguish weeds and crop seedlings has become the most scientific and most effective method. By extracting the effective features of plant images, we can identify and classify plant seedlings very well.

This paper mainly studies the computer image processing technology based on MATLAB to identify plant seedlings. In the process, the image processing toolbox of MATLAB is used. By studying different images of 12 different growth stages of plants, the image is pretreated first, including background separation, contour extraction, morphological processing, and size return. Secondly, the feature parameters are extracted from three aspects of color features, texture features and shape features, and color moments are extracted as color feature parameters, and the energy, entropy, moment of inertia and correlation are extracted as texture parameters by using the gray level co-occurrence matrix, and HOG features are extracted as shape parameters. Finally, HOG is used. Features, combined with the SVM support vector machine classifier, the classifier is trained and the classification of the test images is completed. By adjusting the HOG features, the classification recognition rate is up to 61.7%. HOG features are easily affected by the angle of shooting and the overlap of the blades. The slow training speed of the SVM classifier also affects the classification efficiency, so the next step will be how to better handle the image to reduce the impact of the physical factors on the results and how to optimize the SVM classifier to improve the recognition rate and recognition efficiency.

.Key words：Feature extraction; HOG feature; recognition; support vector machine

剩余内容已隐藏，请支付后下载全文，论文总字数：17413字