In practice, acquiring and annotating data in specialized domains can be costly, thereby constraining the performance and applicability of deep learning. Utilizing generative models to synthesize data proves to be an effective augmentation technique. Therefore, this research proposes a diseased leaf generation pipeline to diversify the maize disease datasets. We introduce the Dual-Perception Cycle-Consistent Generative Adversarial Network (DP-CycleGAN). During training, incorporating our proposed Structure Perception (SP) loss and Texture Perception (TP) loss functions. These losses guide the model 's attention areas through activation reconstruction and mask mechanisms, thereby improving the overall perceptual quality of the generated images and the realism of the disease lesions. We constructed a maize leaf mixed disease dataset to simulate the complex conditions of real-world disease occurrence. Experimental results show that the DP-CycleGAN generates higher-quality and more realistic diseased leaf images. Compared to CycleGAN and state-of-the-art method, DP-CycleGAN shows a 29.6% and 15.7% reduction in Fre chet Inception Distance (FID) scores and a 125.3% and 61.5% increase in Structural Similarity (SSIM) values, respectively. Simultaneously, by incorporating synthetic data during training, our approach significantly enhances the performance of the recognition model in scenarios of both data abundance and scarcity, with improvement rates exceeding two times those of existing state-of-the-art methods. This contributes to the application of Artificial Intelligence (AI) in agricultural production practices.

Rice planthoppers are among the most severe migratory pests affecting rice, characterized by small size and rapid reproduction, leading to sudden and explosive outbreaks. Therefore, timely and accurate monitoring of rice planthopper populations is crucial. Applying machine vision to field monitoring of rice planthoppers can reduce labor and material costs. Existing literature lacks research on field detection and counting of rice planthoppers, and general detection and counting methods suffer from performance degradation in complex environments. In this study, we propose the Rice Planthopper Counter (RPH-Counter), a novel detection and counting architecture. The model is a simple Fully Convolutional Network (FCN). Initially, we propose the Object Counting loss (OC loss), which includes four sub-loss functions that compel the FCN to learn each object's center and boundary positions while constraining false positives. After training, the FCN can predict a separate spot for each rice planthopper, achieving precise localization and counting of the pests. Then, we propose the Self-Attention Feature Pyramid Network (SAFPN) by adding additional Spatial Self-Attention (SSA) modules at the lateral connections of C3 to C5, enhancing the model's performance in complex environments at a lower computational cost. We collected a large-scale field rice planthopper dataset, containing approximately 140,000 annotated rice planthoppers. The evaluation metrics are localization accuracy and counting error. Experimental results show that the RPH-Counter, with lower computational complexity, significantly improves performance, achieving an F1 score of 92.36%, a Mean Absolute Error (MAE) of only 2.40, and an R-squared (R2) of 0.985. Compared to the state-of-the-art object detectors, the F1 score improved by 8.62%, and the counting error decreased by 61%. Compared to the state-of-the-art density estimation methods, the counting error decreased by 23%, with precise localization ability and multi-class expandability. This method offers a new research approach and promising direction for field pest counting and pest population monitoring.

The intramuscular fat content is an essential indicator of pork quality, directly affecting sensory quality and consumers' willingness to buy. Traditional testing methods are subjective and destructive, their assessment is scored by trained assessors according to a marbling scale, but the human sensory evaluation has great subjectivity and randomness. Nowadays, there are more methods to predict the intramuscular fat content of livestock meat using computer vision techniques. However, the complex background of the image makes it difficult to segment the target and background. This study proposes a method based on semantic segmentation networks and machine learning algorithms to detect the intramuscular fat content of multi-part pork cuts. The images of five different pork cuts (belly, loin, collar, ham, and hock) are used as experimental data in the study, the results show that the method proposed in this paper can objectively detect the intramuscular fat content of pork and the average accuracy of prediction can reach 93.28%. This new information method can be used to assess the quality of pork in markets and food processing plants, enabling processors and consumers to distinguish pork cuts with different fat content. It improves the quality of pork to meet the needs of the food processing industry.

Pest management has long been a critical aspect of crop protection. Insect behavior is of great research value as an important indicator for assessing insect characteristics. Currently, insect behavior research is increasingly based on the quantification of behavior. Traditional manual observation and analysis methods can no longer meet the requirements of data volume and observation time. In this paper, we propose a method based on region localization combined with an improved 3D convolutional neural network for six grooming behaviors of Bactrocera minax: head grooming, foreleg grooming, fore-mid leg grooming, mid-hind leg grooming, hind leg grooming, and wing grooming. The overall recognition accuracy reached 93.46%. We compared the results obtained from the detection model with manual observations; the average difference was about 12%. This shows that the model reached a level close to manual observation. Additionally, recognition time using this method is only one-third of that required for manual observation, making it suitable for real-time detection needs. Experimental data demonstrate that this method effectively eliminates the interference caused by the walking behavior of Bactrocera minax, enabling efficient and automated detection of grooming behavior. Consequently, it offers a convenient means of studying pest characteristics in the field of crop protection.

针对当前玉米病害发生量大、病情复杂、难以防治,严重影响玉米产量和质量的问题,提出了一种基于卷积神经网络和迁移学习的玉米叶片病害检测与识别方法.首先收集了 3 827张玉米健康叶片图像和3种不同的玉米病害叶片图像样本,为了使模型拥有更好的泛化能力,使用生成对抗网络对样本进行处理,得到分辨率更高的样本,再对样本进行平移旋转,使样本数量达到5 153张.然后构建ResNet模型,分别对ResNet34、ResNet50及对其添加CBAM注意力机制和FPN特征金字塔网络,并通过迁移学习方法将预训练权重迁移到训练模型中.试验结果表明,ResNet50结合CBAM注意力机制模型的准确率达到了 97.5％,相比ResNet50模型准确率提升了 4.2百分点,相比ResNet34模型准确率提升了 4.9百分点.本研究表明,提出的ResNet50结合CBAM注意力机制模型能够较精准地检测识别玉米枯萎叶、锈病叶、灰斑病叶和健康叶.并可将模型安装在无人机等移动设备上,实现对玉米叶片病害智能化防治,而且后期还会扩充更多的植物病害数据,实现对多类植物病害的检测,为智慧农业添砖加瓦,促进农业防治现代化.

针对现有茶叶病害识别方法准确率低、模型运行速度慢和缺乏检测功能等问题,提出一种新的Yolov7-TSA(Yolov7-Tiny-SiLU-Attention)轻量型网络架构对茶叶病害检测和分类.将Yolov7-T网络中的LeakReLU激活函数替换为SiLU激活函数,以提升检测精度,并防止过拟合.同时,通过融合坐标注意力机制进一步提升对目标轮廓和空间位置的特征感知能力.在含8种茶叶病害(含健康茶叶)的数据集上试验.结果表明,Yolov7-TSA网络对茶叶病害的识别准确率达到了 94.2％,与Yolov7-T、Yolov7网络相比,分别提升了 3.2、1.2个百分点.另外,Yolov7-TSA网络在参数量、浮点运算数、模型大小和单张图片推理时间方面表现出了显著的效果,与Yolov7网络相比,其分别降低了 83％、87％、83％和34％.该网络模型实现了对茶叶病害的检测与分类,同时平衡了识别准确率和实时性能.

Studying the grooming behavior of Bactroceraminax adults at rest and the behavioral interference among multiple adults can provide a research basis for related animal behavior studies. The traditional method of manual recording grooming behavior is time-consuming and error-prone. Based on computer vision and deep learning technology, we first build the improved Yolov5 object detection algorithm to detect B.minax in the video data, and then combine the detector and the improved DeepSort object tracker to track each B.minax individual, and finally build the spatio-temporal feature detection model to recognize and quantify the grooming behavior of each B.minax individual. Using our method to recognize the grooming behavior of a total of 23 B.minax from 4 videos, the results show that the average accuracy rate is over 96%, and the standard deviation is less than 3%. Compared with the existing methods, our method has a higher accuracy rate, and the deviation is controlled within a certain range. Therefore, while significantly improving the quantification efficiency, the method described also guarantees the accuracy of grooming behavior recognition and provides new ideas and methods for studying insect behaviors.

The object detection algorithm is mainly focused on detection in general scenarios, when the same algorithm is applied to drone-captured scenes, and the detection performance of the algorithm will be significantly reduced. Our research found that small objects are the main reason for this phenomenon. In order to verify this finding, we choose the yolov5 model and propose four methods to improve the detection precision of small object based on it. At the same time, considering that the model needs to be small in size, speed fast, low cost and easy to deploy in actual application, therefore, when designing these four methods, we also fully consider the impact of these methods on the detection speed. The model integrating all the improved methods not only greatly improves the detection precision, but also effectively reduces the loss of detection speed. Finally, based on VisDrone-2020, the mAP of our model is increased from 12.7 to 37.66%, and the detection speed is up to 55FPS. It is to outperform the earlier state of the art in detection speed and promote the progress of object detection algorithms on drone platforms.

The reading of the ship draft is an important step in the process of weighing and pricing. The traditional detection method is time-consuming and labor-consuming, and it is easy to lead to misdetection. In order to solve the above problems, this paper introduces the computer image processing technology based on deep learning, and the specific process is divided into three steps: first, the video sampling is carried out by the UAV to obtain a large number of pictures of the ship draft reading face, and the images are preprocessed; then, the deep learning target detection algorithm of improved YOLOv3 is used to process the images to predict the position of the waterline and identify the draft characters; finally, the prediction results are analyzed and processed to obtain the final reading results. The experimental results show that the ship draft reading method proposed in this paper has obvious effects. The method has a good detection effect on high-quality images, and the accuracy rate can reach 98%. The accuracy rate can also reach 73% for the images with poor quality caused by improper capture, character corrosion, bad weather, etc. This method is a kind of artificial intelligence method with safe measurement process, high measurement effect, and accuracy, providing a new idea for related research.

Using Edge Computing and face recognition technology to build intelligent security system is an important research content in modern artificial intelligence application field. In this paper, an insensitive SMS notification system of multi-face recognition is constructed, which collects real-time face recognition data, then analyzes the face information, and sends the SMS to the specified person according to the analysis result. The system in this paper can not only be used for independent short message notification function, but also can be integrated into other traditional face recognition security system.