Maize is an important food crop both in the world, and its breeding work holds great significance. This paper conducts research in response to the requirements of three-dimensional (3D) seed testing for maize ears. Fifty maize ears with different morphologies were selected as samples. A 3D scanning device was used to obtain 3D point cloud data, and a textured 3D reconstruction model was obtained after processing. Preprocessing was carried out on the acquired point cloud data. Firstly, the Iterative Closest Point (ICP) algorithm was employed for point cloud registration to transform the point cloud data from different perspectives into the same coordinate system. Secondly, voxel filtering was used for point cloud denoising and smoothing to remove noise. A point cloud segmentation method based on the region-growing algorithm and the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm were used to segment and cluster the maize kernels on the ears. The segmentation results of the maize kernels were obvious, and the Root Mean Square Error of the kernel count of the maize ear samples was 4.25, which was relatively close to the actual kernel count measured manually.

With the increasing popularity of the Internet and intelligent communication devices, communication means and information access have become more diversified, which has also led to the frequent occurrence of cyber fraud. In order to reduce the occurrence of cyber fraud and conduct in-depth research on the characteristics of cyber fraud, this paper collects the criminal judgment documents of the first instance of cyber fraud cases, analyzes the extractable information and extracts the information, visualize and analyzes the extracted information, and obtains the characteristics of cyber fraud cases via visual analysis. The extracted information is visually processed, and the characteristics of cyber fraud cases are analyzed as follows: in cyber fraud cases, the defendants are mostly male, most of them are between 20 and 30 years old, most of them are junior high school educated, and most of them are unemployed; The locations of the cases are mainly distributed in the southern coastal provinces and central China. The trend of team crime is obvious, and the amount involved is large, generally more than ten thousand Yuan; the main types of fraud cases are profit inducement.

Real-time and accurate estimation of maize grain kernel yield enables farmers to make corresponding management decisions, maximize yield and increase profits. In the past, manual calculation of kernels number was not only cumbersome, labor-intensive and slow, but also caused many errors due to the limitations of manual operations. In the kernel detection application, the number of targets to be detected in a single image is large and the targets are adjacent, so it is very challenging. In order to improve the efficiency of corn ear counting, this paper proposes a One-Stage target detection algorithm based on RetinaNet. First, build a feature pyramid network based on ResNet50. Second, use a priori box to improve prediction performance and Focal Loss to balance positive and negative samples. By optimizing the design of the anchor box, the algorithm can be applied to the detection of a large number of small targets, thereby improve the accuracy. Finally, the algorithm is tested on the data set, and the results show that the algorithm can not only successfully label and count all the maize kernels on the ear instead of off the ear.

A Unet model is built segment the different parts of maize ear. However, the accuracy does not meet the needs of application. In the paper, a new activation function named R_S is defined on the basis of the analysis and research on two commonly used activation functions ReLU and Swish. And the new activation function combined the advantages of these two functions and effectively avoided the problems of hard saturation, mean shift and complicated calculations. The U-Net deep learning model with R_S as the activation function, which is used for image segmentation of maize ears, not only had fast convergence speed and high accuracy, but also avoided local convergence effectively. Next, in order to determine the optimal value of adjustment factor, seven U-Net models are built. The adjustment factors are separately 1.7,1.3,1.0,0.8,0.6,0.3,0.05. Experiment results showed that the model could reach the best comprehensive performance when using the self-defined activation function on condition that the adjustment factor is 0.8. Finally, the optimal model is used as the classifier to build the automatic maize ear image segmentation method. The experiment results demonstrate that the accuracy of the model on test set is 94.34%.

The image segmentation for maize ears is a key step in the process of automatic measurement of maize phenotypic parameters. The accuracy of image segmentation has a direct impact on the precision of subsequent calculation on phenotype parameters of maize ear. A segmentation model of maize ear image was proposed to solve the problem of the unclear boundary between adjacent regions in maize ear images. The model realized the standard U-Net model based on the encoder-decoder principle and could integrate feature selection, feature extraction and feature classification. We build two different models with fixed learning rate and stepwise decay learning rate separately. The experimental results demonstrate that using the stepwise decay learning rate in the training process can not only obtain a faster convergence speed but also ensure better convergence value. Overfitting is a common problem in the deep learning model. We adopt regularization and dropout strategies to suppress overfitting. The experimental results show that the regularization strategy can successfully suppress the overfitting problem in this model. The experiment results showed that the accuracy of the model on training set was 93.91% and on test set was 93.45%. The accuracies on each data set were very close, which showed that the model had effective generalization performance.

教育部提出新工科战略规划,为新一轮高等教育改革指明了目标和方向,同时对高等教育质量和大学生的个性化培养提出了更高的要求.“以学为本”的理念基于知识建构主义理论,为教学效果的提升提供了科学基础.以“Python程序设计”课程为例,分析了“以学为本”理念的实施不能仅仅依赖于利用互联网和多媒体技术对授课模式进行变革,更需要在授课环境的设置、教学目标的确定、知识体系的构建、教学逻辑的设计等方面进行全过程、全方位的改革,从而保障“以学为本”理念得以有效实施,培养学生适应现代社会的快速变化和未来不确定的变革挑战.

In this study, we propose a method for correcting the adverse effects produced by the cur vature of fruit objects in images acquired by cameras in machine vision systems. The areas near the edge are darker in acquired images than in the centre, which results in many difficulties for subsequent analyses. In this paper, the fruit object was considered as a Lambertian surface. The light intensity was analysed and the height and normal on fruit surface was deduced based on the shape-from-shading (SFS) algorithm. The geometric correction factors were calculated and the adverse effects of light intensity were corrected on the fruit surface. The proposed method was evaluated on a test set of four types of fruit. The results show that the non-uniformity of the greyscale value on the fruit surface fell by 35.5% after correction, and the ref lectance in the central area of fruit is similar to that of the peripheral areas when using the proposed method. The experiments prove that our method allowed homogenisation of the greyscale level of the pixels belonging to the same class, regardless of where they are on the fruit surface, which will facilitate subsequent classification tasks.

为提高玉米果穗发育程度检测的自动化程度与精度,提出一种基于机器视觉技术的测量方法.在随机森林机器学习算法的基础上构造秃尖、干瘪和籽粒区域的识别模型.该模型由多个独立同分布的弱分类器构成,对输入的训练样本进行列和行两个方向上的随机采样.比较随机森林模型和决策树模型的分类效果可知随机森林模型有效避免了过拟合和局部收敛现象的产生,并具有良好的推广能力.为确定最优的弱分类器数目,选择弱分类器个数为训练样本数量的1/80、1/40、1/20、1/10、1/5、1/4时分别构建随机森林分类器.研究结果表明,当随机森林中弱分类器个数为训练样本数量的1/20时,模型的识别率与稳定性最好.然后,以最优的随机森林模型作为分类器构建玉米果穗不同发育程度自动检测方法.试验结果表明,各区域长度测量的准确性均在95％以上,测量速度可达30个/min以上.

从美国引入的翻转课堂教学模式为改善目前大学生的学习倦怠现象,激发学生学习热情与主动性带来了希望.各高校将工作重点放在教学资源建设与相关师资培训上,却忽视了该教学模式的主体即影响翻转课堂教学模式实施的关键角色——学生.中国学生的课堂参与度低是该模式成功实施的主要障碍,而其根源在于历史和文化两个方面.中美的不同历史导致了不同的教育制度,中美文化差异造就了不同的民族性格.因此,在美国课堂上得以顺利实施的翻转课堂如果要成功移植到中国大学的课堂上,还需采取各种有效措施克服与矫正历史与文化产生的负面影响.

In order to explore a rapid and non-destructive method for identifying storable quality of indica paddy,experiments were conducted on 80 samples of severely unstorable indica paddy and 80 samples of other indica paddy.Support vector machine(SVM) method was selected to build qualitative model according to training samples’characteristics.A combined kernel function was defined after analyzing and studying different kernel function.The proposed combined kernel function was the linear combination of polynomial function and radial basis function,and it combined each advantage of these two functions to acquire both interpolation and extrapolation properties of kernel function.The experiments showed that the model had the best comprehensive performance when the linear combination polynomial function and radial basis function was taken as the kernel function of SVM,and the control factor was 0.7.The training identification rate was 97.21%,and the test identification rate was 93.25%.

In the application of near-infrared reflectance spectroscopy (NIRS) to estimate amylose content of rice, analysis was performed based on the relationship between absorbance and amylose content. But absorbance is sensitive to environmental temperature. In this study, characteristics of rice spectra collected at different temperatures were studied by comparing the average spectra and standard deviation spectra of rice with same amylose content at the temperature of 5, 10, 15 and 20°C. The results showed that spectra collected at different temperatures had significant difference in absorbance, and spectra acquired at different temperatures had different stabilities. When the environment temperature was 15°C, stability of spectra was better than when the temperature was 5, 10 and 20°C within the range consisting of abundant information. Accuracies of four linear models on calibration set acquired at four temperatures built by partial least square (PLS) were investigated by using validation sets got at four different temperatures to perform cross validation. The results showed that the relation between spectra and amylose content was not linear in the result of temperature’s influence, and non-linear model methodology least square support vector machine (LS-SVM) method was used to establish calibration model on the calibration set acquired at 15°C. The validation results indicated that the non-linear model could predict the amylose content of samples collected at 15°C with high accuracy, and the root mean square error of prediction was 0.62. Meanwhile, non-linear model showed good prediction ability for samples whose spectra were collected at other temperatures. Key words: Amylose content, near-infrared spectroscopy, temperature influence.

Credit risk had become the main cause of bank bankruptcy. Therefore, risk control and risk management became the most important problem of credit loan. This study used a series of credit evaluation indices to quantify the credit of farmer in the form of number. Two models were built on these indices according to the association rules and cloud theory. Model A was used when the customer loaned from credit cooperative at the first time. Model B was used when the farmer was a regular customer who had loan records in the credit cooperative. This system provided credit cooperative with important credit loan data, and made decision-making more scientifically and closer to reality, thus evade credit risk effectively.

In the process of collecting spectrum using near-infrared instrument, the final spectrum signal detected by detector contains not only sample related information but also the noise signal because of influence from various aspects of interference factors. In order to eliminate interference resulted by drift and shift of baseline, random noise and stray light of instrument, and to improve signal-to-noise ratio (SNR) of spectrum, various baseline correction methods and spectral smoothing methods were used to process rice spectrums. And then eight models were established. The model which used first derivative and Savizky-Golay convolution had the best performance with comparison of various performance indicators including correlation coefficient(R), root mean square error of calibration(RMSEC), root mean square error of cross validation(RMSECV).

提升农村金融服务水平是我国农村金融领域改革的热点和难点问题,也是农村金融改革的最终目标之一.2009年"中央一号"文件指出要提高农村金融服务水平质量和水平.

In order to optimize calibration set and increase prediction accuracy of the calibration model when near infrared spectroscopy was used to develop the model for rice amylose content, 18 abnormal spectrums produced by subjective and objective factors were eliminated based on Mahalanobis distance criterion combined with prediction concentration residual standard. The calibration results showed that the correlation coefficient of calibration model increased from 0.86287 to 0.9350, and root mean square error of calibration reduced from 2.53 to 1.54. The correlation coefficient of cross validation using Leave-One-Out method increased from 0.62785 to 0.86850, and root mean square error of cross validation reduced from 4.05 to 2.18.

针对传统轮廓提取方法自适应能力不强的问题,提出一种基于蚁群算法的稻种轮廓提取方法.将数字图像看作一个二维网格,以稻种轮廓为蚂蚁要寻找的食物,以梯度与方向作为启发信息,蚂蚁在启发信息与信息素强度的共同引导下搜索真实轮廓.为了加快算法的收敛速度,增强蚁群的搜索能力,定义了合理的启发信息与信息素更新机制,并研究了算法中初始参数之间的关系与规律.最后,用两种光照条件下采集的图像测试该算法.实验结果表明,该算法具有并行性、鲁棒性,自适应能力较强,提取的轮廓较准确.

Farmers are the main financing of rural financial markets, in rural financial institution, lending is based on the credit status of the farmers, so the evaluation of their credit level is always the focus of the theory field and the practical field. A reasonable credit evaluation index system is the prerequisite for scientific credit evaluation. In this study, association rules technology which is the core methods of data mining was applied to mine credit database, and the weight of each index and the score of each attribute value could be accurately calculated by the results of correlation analysis. Through this way, the establishment of credit evaluation index system was more scientific, objective and reasonable.

A method based on the cloud theory was developed to improve the automatic degree and accuracy of chalkiness detection. In this method, without man's intervention, chalkiness and non-chalkiness were defined as two qualitative concepts. An asymmetrical cloud was used to represent chalkiness, and a symmetrical cloud was used to represent non-chalkiness. These two clouds were respectively described by two groups of digital characters. Firstly, dynamic threshold program was designed to acquire training samples for the two clouds. Secondly, backward cloud generators were developed to implement the transformation from the quantities to the qualitatives. Finally, maximum value judgment method was used to separate the chalky region from non-chalky region according to the membership function of each cloud. The result shows that the classification accuracy of cloud classifier is higher than the classification accuracy of traditional hard classifiers.

农村信用社农户小额信贷是我国小额信贷的主流,此类贷款是信用社以农户的信誉为保证。在对湖北省6个具有代表性的传统农业县市进行入户调查的基础上,将原始数据经过预处理后,用概率神经网络建立预测模型,从而实现对农户信用的快速评价,以此作为信用社放款与否的依据。

The automatic detection of grain is the important application of digital image processing and pattern recognition technology.The process of this application includes several steps.Firstly,the acquired image of grain is preprocessed.Sec-ondly,the typical features are extracted from the processed grain image.Finally,the selected features are treated as the in-put of classifier after being analyzed.The recent progress of machine vision technology in grain recognition was summarized in the light of image processing,image analysis and image recognition,and the application of these approaches was also re-vealed.