To address the challenges of significant scale variations and difficulty in extracting discriminative features for small targets in airborne remote sensing object detection, while considering the constraints and efficiency requirements of onboard systems, this article proposes a lightweight detection algorithm. First, a structural reparameterization strategy is applied to optimize depthwise separable convolutions, simplifying complex structures during the inference stage, which improves inference speed and memory utilization. Second, the G-Shuffle module is designed to significantly enhance feature extraction efficiency and interchannel information interaction, balancing computational complexity and detection accuracy. Lastly, a dynamic multiscale pyramid network is introduced, employing pyramid convolution to effectively extract and fuse multiscale features. Additionally, a channel attention network is integrated to establish channel interconnections, expanding the receptive field and enhancing the extraction of finegrained features at different scales, thereby improving the detection capability for targets of various sizes, especially small targets. Experimental results demonstrate that the proposed method achieves a mAP@0.5 of 0.552 and mAP@0.5:0.95 of 0.265 on the AI-TOD dataset, with a 2.7 ms improvement in inference speed and a 14% reduction in model parameters. These results indicate superior performance across multiple metrics compared to existing state-of-the-art methods, proving the method's potential and practicality for airborne remote sensing object detection.

The impact of porous structures on the combustion characteristics and energy efficiency of premixed NH3/H2 fueled micro-combustors for micro-scale thermophotovoltaic and thermoelectric applications is investigated. Three triply periodic minimal surface (TPMS) structures, including diamond, IWP, and primitive structures, are analyzed with a focus on gradient porosity under different operating conditions. Structural optimization significantly enhances fuel combustion efficiency and burner radiation power. The optimized gradient diamond TPMS combustor (C24) with proper materials of combustor and porous media demonstrates superior performance, achieving the maximum outer wall temperature of 1171.61 K with high combustion efficiency at input energy of 100 W. A sensitivity analysis is conducted to explore the combustion characteristics under different fuel blending ratios. Structural optimization, particularly in the combustor with a diamond structure, enhances residence time, improves heat transfer, and stabilizes the flame, achieving a peak radiation efficiency of 32.94 %. This represents a 25.28 % improvement compared to the empty combustor. A novel modeling and simulation approach, providing comprehensive analysis of reaction processes, flow fields, temperature distributions, and chemical reaction kinetics is introduced, which offers valuable insights for optimizing micro-combustor designs through parameter refinement.

Rapid and nondestructive identification of metabolites in fresh tobacco leaves is crucial for understanding quality formation and optimizing production. This study utilized hyperspectral imaging (400-1000 nm) to analyze fresh tobacco leaves at three maturity stages (M1, M2, and M3) and identified 11 key metabolites with significant differences. Four machine learning models-Random Forest (RF), extreme gradient boosting (XGB), convolutional neural network (CNN), and long short-term memory (LSTM)-were employed to predict metabolite content. Experimental results demonstrated that CNN achieved the highest prediction accuracy, with an average R2 of 0.917, outperforming RF (0.896), XGB (0.895), and LSTM (0.915). The superior performance of CNN highlights the potential of deep learning in metabolomics research. These findings provide insights into the metabolic dynamics of tobacco maturation and offer practical applications for quality control and production optimization in the industry.

Thermoelectric generator (TEG) technology presents a highly attractive solution for converting waste heat recovery. However, the low energy efficiency limits the application, and researches predominantly focuses on component enhancements, the system-level coordinated optimization is substantially required. To address the gap and improve the energy conversion of waste heat recovery, a H2/NH3 fueled burner incorporating integrated blocks and ribs to augment heat transfer, coupled with two-stage TEG for enhanced thermoelectric conversion is proposed. The results indicate that increasing reactants flow rate initially elevates the TEG power output, while further increments reduce system efficiency. Combustion stability and power generation performance are significantly influenced by the reactants equivalence ratio h and fuel properties, where optimal performance is observed at h = 0.9 and NH3 blending ratio gamma = 30 %. Furthermore, burner thermodynamic behavior is strongly dependent on its heat dissipation characteristics, leading to an optimal fin geometry with fin height hf = 14 mm and width wf = 0.5 mm. The incorporation of blocks and ribs in combustion chamber is demonstrated to effectively enhance energy conversion and heat transfer, thereby elevating the power output. The proposed twostage TEG significantly increases the maximum output power and extends the effective operational range of the system. At reactants mass flow rate mr = 3.226 x 10-5 kg/s, the optimized thermoelectric system achieves an output power 11.75 % higher than that of the conventional single-stage TEG.

Speckle noise is a significant challenge in synthetic aperture radar (SAR) images, severely degrading the visual quality and compromising subsequent image interpretation tasks. While existing despeckling methods can reduce noise, they often fail to strike a appropriate balance between noise suppression and the preservation of fine image details. To address this issue, in this paper, we propose a novel SAR image despeckling method that leverages both structural image priors and noise distribution characteristics in an end-to-end framework. Our approach consists of two key components: a dual-branch subnet for coarse despeckling and noise estimation, and a noise-guided Transformer-based subnet for final image refinement. The dual-branch subnet decouples the tasks of noise estimation and despeckling, improving both noise suppression accuracy and structural detail preservation. Furthermore, a combination of grouped pooling attention (GPA) and context-aware fusion (CAF) modules enables effective multi-scale feature fusion by jointly capturing local details and global contextual information. The noise estimation branch generates adaptive priors that guide the Transformer refinement, which incorporates deformable convolutions and a masked self-attention mechanism to selectively focus on relevant image regions. Extensive experiments conducted on both synthetic and real SAR datasets demonstrate that the proposed method consistently outperforms current state-of-the-art methods, achieving superior speckle suppression while preserving fine details more effectively.

Elemental composition design is a common strategy for tuning the crystallographic structure of films, thereby optimizing their mechanical properties. The evolution of film structure is highly dependent on both selection and concentration of cation and anion elements. In this work, we utilized strong nitride-forming elements (Hf, Nb, Zr, and Ta) and the weak nitride-forming element Mo as the metallic cations in a multi-principal element nitride system. The nitrogen content in the (HfMoNbZrTa)1-xNx film was manipulated by varying the N2/Ar flow ratio (RN) during reactive magnetron sputtering. The results showed that the HfMoNbZrTa metallic film crystallized weakly into a body-centered cubic (BCC) structure, yielding a hardness (H) of approximately 10.7 +/- 0.2 GPa. The incorporation of nitrogen immediately induced a transformation from BCC to a face-centered cubic (FCC) predominated multi-phase structure at substoichiometric regimes (10% <= R N <= 20%). A further increase of R N to >= 25%, i.e., at near-stoichiometric regimes, resulted in the formation of single-phase FCC structure. Hardness and wear rates both reached their optimal values at R N = 20-25%. The strengthening mechanism was elucidated through density functional theory (DFT) calculations, which suggest that higher H of the substoichiometric film is mainly associated with the increased local lattice distortion and the formation of a multi-phase structure.

Visual detection for automated fruit harvesting in unstructured environments constitutes a critical technical challenge, especially for fruit peduncles, which exhibit greater sensitivity to environmental factors than the fruits themselves. To address this challenge, this paper proposes a top-down keypoint detection method for pepper peduncles in unstructured environments. The proposed method enables accurate estimation of peduncle poses. The first step of the research involves validating different object detection models and employing ones to identify the bounding boxes of pepper peduncles. Subsequently, a new keypoint detection model based on the Lite Vision Transformer is proposed, leveraging the Transformer's capacity to capture long-range spatial and semantic dependencies. Experimental results on the pepper dataset collected in unstructured environments demonstrate that the proposed model achieves an AP50 of 94.6 %. This performance surpasses multiple state-of-the-art keypoint detection methods while maintaining lightweight parameters and low computational complexity. Moreover, a series of tests reveals that the proposed method outperforms other algorithms in complex environments, especially in occlusion scenarios. Finally, a comprehensive evaluation of the top-down approach is conducted, examining the influence of object detection and keypoint detection models on overall performance. The proposed keypoint detection model achieves the highest performance, with a detection speed of 9.38 FPS when using YOLOv8s as the object detection model, and an AP50 of 83.6 % when using YOLOv8l. Experiments conducted in real unstructured environments demonstrated the robustness of the proposed method, effectively detecting the posture of dense and occluded chili pepper peduncles. This research can be extended to the detection of fruit peduncles in other crops, providing a foundation for pose estimation of fruit peduncles in complex environments.

Solar cell defect detection is crucial for quality inspection in photovoltaic power generation modules. In the production process, defect samples occur infrequently and exhibit random shapes and sizes, which makes it challenging to collect defective samples. Additionally, the complex surface background of polysilicon cell wafers complicates the accurate identification and localization of defective regions. This paper proposes a novel Lightweight Multiscale Feature Fusion network (LMFF) to address these challenges. The network comprises a feature extraction network, a multi-scale feature fusion module (MFF), and a segmentation network. Specifically, a feature extraction network is proposed to obtain multi-scale feature outputs, and a multi-scale feature fusion module (MFF) is used to fuse multi-scale feature information effectively. In order to capture finer-grained multi-scale information from the fusion features, we propose a multi-scale attention module (MSA) in the segmentation network to enhance the network's ability for small target detection. Moreover, depthwise separable convolutions are introduced to construct depthwise separable residual blocks (DSR) to reduce the model's parameter number. Finally, to validate the proposed method's defect segmentation and localization performance, we constructed three solar cell defect detection datasets: SolarCells, SolarCells-S, and PVEL-S. SolarCells and SolarCells-S are monocrystalline silicon datasets, and PVEL-S is a polycrystalline silicon dataset. Experimental results show that the IOU of our method on these three datasets can reach 68.5%, 51.0%, and 92.7%, respectively, and the F1-Score can reach 81.3%, 67.5%, and 96.2%, respectively, which surpasses other commonly used methods and verifies the effectiveness of our LMFF network.

Accurately measuring tobacco harvest maturity is crucial for optimizing quality and crop management. Traditional methods heavily rely on qualitative evaluation and chemical experiments, which introduce subjectivity and inherent limitations. This study proposes a novel method that combines deep learning with proximal hyperspectral imaging (HSI) to achieve precise recognition of tobacco leaf maturity. Unlike traditional techniques, HSI captures rich spectral and spatial data, enabling comprehensive analysis. It adopts a pixel-level annotation strategy to annotate each pixel based on its maturity level, thereby preserving spectral complexity. The dataset comprises 3000 randomly extracted cubes (5x5x176) from 150 original hyperspectral images, encompassing five maturity levels. Spectral and spatial features are extracted from hyperspectral data using a three-dimensional convolutional neural network (3D-CNN) architecture. This method effectively leverages complex spectral patterns for maturity recognition. During testing, the model demonstrated an impressive average accuracy of 99.93%. Visual predictions vividly illustrate the model's proficiency in maturity recognition, affirming its practical utility. This study pioneers the integration of deep learning and hyperspectral near-end sensing technology in tobacco maturity assessment, mitigating the constraints of traditional methods, and establishing the groundwork for real-time monitoring and quality control of tobacco.

Due to the economic differences of chili peppers at different maturity levels and the absence of differentiation of maturity in existing harvesting processes, it is crucial to accurately identify the maturity during harvesting for improving economic benefits. This paper investigated pepper maturity recognition models using hyperspectral technology to realize intelligent pepper harvesting with high accuracy and monitor maturity. Hyperspectral data (400-1000 nm) for field line peppers were collected, preprocessed using normalization, Savitzky-Golay convolutional smoothing, and standard normal variable transformation and subsequently used to train back propagation neural network (BP) and kernel based extreme learning machine (KELM) models. Principal component analysis (PCA) was used to reduce data dimensionality and identify characteristic spectral wavelengths for pepper maturity at 862.2, 676.9, 578.1, and 980.7 nm; and BP, PCA-BP, KELM, and PCA-KELM models were subsequently established. The precision for the KELM and PCA-KELM models (99.5% and 97.3%, respectively) was superior to the other two models; however, the PCA-KELM model used only four of the feature wavelengths as inputs, hence it required only 1/44 the data compared with the KELM model. Thus, the PCA-KELM model achieved high recognition accuracy and training speed, offering an effective method to discriminate pepper ripeness based on hyperspectral features.

Deep neural networks have received wide applications in many areas in recent years, but the high requirements of computational cost and storage space limit its applications in embedded devices. To deal with this problem, we proposed a lightweight re-parameterization YOLOv5 network with four detection heads, called RepFYOLOv5. First, a structure re-parameterization module RepBlock is proposed to reconstruct the network backbone, which has multiple topology branches during the training process, but in the inference and deployment, it uses the equivalent convolutional module fusion structure to speed up the network inference. In addition, in order to improve the object detection accuracy (especially for the small objects), except for the original three detection heads of YOLOv5, an extra detection head based on larger feature maps is added to the network to improve the detection accuracy for the small objects. Experimental results on the COCO dataset and our own traffic surveillance dataset showed that our proposed lightweight RepFYOLOv5 obtained a better balance among detection accuracy, inference time and parameter amount etc.

This study centered around the practical problem that there is no machine available for deep planting with large holes in hilly and mountainous areas of China. According to the principle of spiral lifting, a conical, double-spiral hole-forming machine was innovatively designed. The structural design and parameter calculation were completed. By discrete-element-method (DEM) simulation, the optimal lead and rotation speed of the hole former were obtained, and the hole-forming mechanisms of soil cutting, soil lifting, soil discharging, soil extruding, and soil returning were further revealed. The field test results indicated that the prototype had the advantages of convenient operation and good performance, and the formed holes met the agronomic requirements, with a qualification rate of 88.5%. In addition, it was found that the soil moisture content has a great influence on the formation of holes. Under the condition of low moisture content, the residence time at the bottom of a hole should be appropriately increased to improve the qualification rate of the holes formed. Our research results provided theoretical guidance and technical support for the design, optimization, popularization, and application of a hole-forming machine for deep planting with large holes (DPLH).

In this paper, three new important aspects of rotary electromagnetic vibration energy harvesting technology (RE-VEH) are concerned and investigated: (i) vibro-electric coupling mechanism of the RE-VEH system is studied through theoretical modeling; (ii) quantitative analysis of system parameters based on numerical simulation method is carried out for the optimal design of RE-VEH; and (iii) dynamic power output performance of the RE-VEH system in free vibration is discussed. The parameter adjusting methods of the RE-VEH system in free vibration mode are obtained through theoretical analysis and numerical simulation. The experimental results show that the power output performance of RE-VEH in free vibration mode matches the numerical simulation results. The simulation and experimental results show that the maximum voltage output and power output of the RE-VEH with different structure parameters under free vibration can be up to the level of 10(0)similar to 10(1) V/watt. The above results indicate that RE-VEH in a free vibration environment has significant energy output performance.

The engineering training center in universities is an important carrier for cultivating students'scientific and technological innovation ability,practical ability and engineering background education.The level of management efficiency of engineering training center is directly related to the utilization level of training equipment and the level of teaching services.Based on the actual situation of the engineering training center of Guizhou University,the existing problems in its management were analyzed and the management efficiency of universities was explored and optimized from the management of students'open reservations,the management of teaching staff,the creation of competition community workshops,the creation of equipment libraries,the improvement of teaching information management,and the strengthening of student participation management,in order to continuously improve the management level of engineering training centers,better serve students,and help cultivate high-level innovative talents.

为了给白萝卜收获机的设计提供必要的数据支撑,对白萝卜缨的物理力学特性进行试验研究.为此,测量白萝卜的茎秆长度、根部长度、茎秆根数、含水率等物理几何特征,研究了白萝卜茎秆直径、含水率、加载速度3个因素分别对于白萝卜茎秆拉伸力的影响;利用Design Expert软件建立了3个因素与白萝卜茎秆直径拉伸特性之间的影响回归模型,分析双因素交互作用对于茎杆拉伸力的影响.结果表明:茎秆直径对于白萝卜茎秆拉伸力影响显著,含水率和加载速度对其影响不显著;得到的回归模型与试验结果拟合程度较好,可用于预测白萝卜茎秆拉伸力的变化情况.试验研究为白萝卜收获机关键部件的设计和研发提供了数据和理论依据支持.

针对白萝卜种子颗粒本征参数、颗粒与种植机械装备间接触参数缺乏等问题,以白萝卜种子颗粒为研究对象,利用三维扫描逆向建模技术与EDEM软件建立白萝卜种子颗粒离散元模型,通过物理试验与虚拟仿真试验对仿真参数进行标定.采用碰撞弹跳试验、斜面滑移试验和斜面滚动试验确定白萝卜种子颗粒与ABS塑料、不锈钢板、有机玻璃和铝合金4种不同材料之间的碰撞恢复系数、静摩擦系数和滚动摩擦系数,碰撞恢复系数分别为0.48、0.62、0.51、0.44,静摩擦系数分别为0.50、0.42、0.45、0.48,滚动摩擦系数分别为0.014、0.025、0.007、0.006;通过响应曲面和Design-Expert软件的优化模块对多元二次方程进行多目标优化,获得白萝卜种子间离散元模型接触参数较优组合:碰撞恢复系数0.19、静摩擦系数0.54、滚动摩擦系数0.02.采用圆筒提升法进行白萝卜种子颗粒物理休止角堆积试验,利用 MATLAB对堆积图像处理获得物理堆积试验白萝卜种子颗粒与ABS塑料、不锈钢板、有机玻璃和铝合金的休止角分别为33.48°、32.72°、33.81°、29.88°,仿真试验与物理试验得到的休止角误差分别为1.4%、3.2%、2.6%和2.8%.研究结果表明,白萝卜种子颗粒建模和标定所得的离散元仿真参数具有准确可靠性,可为白萝卜种子颗粒离散元仿真研究提供理论参考.

烤烟油分含量评价对烟叶等级的评判具有重要影响,为了对烤烟油分等级进行科学预测,创建烤烟微观纹理与油分含量的关系模型,论文以贵州安顺平坝烟区烤烟样品为研究对象,基于LBP—GLCM(融合灰度共生矩阵与局部二进制模式)特征融合提取烤烟表面纹理特征,结合BP人工神经网络模型,对烤烟油分等级进行预测.结果表明:采用LBP-GL-CM算法结合BP神经网络对烤烟油分等级预测正确识别率为93.33%,模型相关系数为0.91486,可见该算法对于烤烟油分等级预测具有一定的优势.

为实现烟叶快速准确的识别分级,提出了一种改进的Faster R-CNN分级算法.以VGG16网络为训练原型,通过调整训练图像的尺寸、学习率、mini-batch等全连接层参数,在此基础上将ROI pooling改进为ROI align,再去掉网络模型的第8、第12、第15层卷积层,同时引入Inception结构层为研究的最终分级模型.以准确率和召回率作为分级模型性能的评价指标,利用最终分级模型训练7个等级的烟叶图像,识别准确率最低为90.62%,最高为92.46%,召回率最低为91.72%,最高为92.87%,平均识别准确率达到92.35%,识别速度达到0.2s/幅.改进后的模型平均识别准确率相比原始网络平均识别准确率提高了3.98%.

In this paper,AT89C51 single chip microcomputer was used as the controller to design a set of water injection system suitable for tobacco water retaining agent in hilly tobacco area.The whole water injection system was mainly composed of the whole machine frame,power component,automatic pipe collecting and releasing mechanism,pipe winding mechanism,submerged water injection gun and water retaining agent water injection operation control system.The system has the characteristics of fast response,high precision and simple operation,which is suitable for the water injection operation of tobacco water retaining agent in hilly tobacco area.

从单目视觉导航和双目视觉导航两方面进行阐述,总结了视觉导航关键步骤图像预处理与导航路径现有提取方法面临的挑战.概括了国内外改进方法,国内研究聚焦于提高导航线拟合算法的实时性和稳定性,国外则更关注增强图像预处理对作物特征提取的效果.针对农业机械视觉导航面临田端换行和实时性挑战,提出了 2 种相应解决思路.最后,阐明了农业机械视觉导航技术将朝着多传感器信息融合导航和自主避障等方面发展.