基于 CNN 算法的缺秧与漂秧图像识别技术研究

插秧机是现代农业向自动化方向发展过程中使用的重要工具之一，由于受到地理环境和设备等因素影响，插秧机在工作中难免会出现缺秧及漂秧等情况。传统对缺秧和漂秧的识别主要依靠经验与人工作业，效率低下、准确度不高，因此提出基于深度卷积神经网络（CNN）算法的缺秧与漂秧图像识别技术。首先计算缺秧与漂秧数据图像样本的质心位置，根据质心间距离是否在合理范围内识别缺秧，然后提取秧苗样本特征建立样本库，对采集的秧苗图像数据进行分析处理，再与样本库进行对比，以此判断插秧机在工作过程中是否存在缺秧和漂秧情况。通过对仿真算例进行测试，验证了算法的有效性，其准确率达到 90%以上。该方法对于农业自动化的发展具有重要意义，对于相关实践能起到一定的推动作用。

Image Recognition Technology of Seedling-lacking and Drifting Seedlings Based on CNN Algorithms

Transplanter is an important means in modern agricultural automation development. Due to the influence of geographical environment and equipment，the rice transplanter will inevitably suffer from lack of seedlings and floating seedlings. Traditional recognition of missing and floating seedlings mainly relies on experience and manual work，which is inefficient and inaccurate. Therefore，this paper proposes a recognition technology of missing and floating seedlings image based on deep convolution neural network（CNN）algorithm. Firstly，the centroid position of missing and floating seedlings data image samples are calculated to identify missing seedlings according to whether the distance between centroids is within a reasonable range，and then the characteristics of seedling samples are extracted to establish a sample database，and the collected image data are analyzed and processed. The sample databases are compared to determine whether the transplanter is lack of seedlings and floating seedlings in the process of work. This paper tests the simulation examples to verify the effectiveness of the algorithm，and its accuracy reaches more than 90%，which proves that this method is of great significance to the development of agricultural automation，and can play a certain role in promoting the relevant practice.