Feeding fish based on their feeding behavior and intensity facilitates their healthy growth and improves aqua-culture efficiency. Current methods assess feeding intensity and behavior based on feeding sounds and splashes. However, evaluating the feeding behavior of juvenile fish is difficult because of their slow feeding rate and small size. This paper presents an improved YOLOv8-feed key points in the fish feeding behavior model (YOLOv8-FB), based on YOLOv8n-pose, to detect juvenile fish feeding behavior effectively. YOLOv8-FB combines target detection and key point recognition; it replaces the bottleneck in C2f with the feed pellets edge information extraction module stem (FPEIEMStem) block, resulting in C2f-FPEIEM. Additionally, the global-to-local spatial aggregation (GLSA) module is integrated into a bi-directional feature pyramid network (BiFPN) to replace the original path aggregation network as a neck network, enhancing feature fusion. FPEIEMStem block leverages a SobelConv branch for edge information extraction and a convolutional branch for spatial information extraction, enabling efficient feature learning that incorporates rich edge and spatial information. Experimental results reveal that YOLOv8-FB achieves a mean average precision of 94.8 % for detecting juvenile fish feeding behavior, reduces the number of parameters by 37.1 %, and has a model complexity of 6.9 GFLOPs. Moreover, a short-term starvation test demonstrated increased feeding activity among juvenile fish, reflected in a higher frequency of contact with feed pellets. YOLOv8-FB has excellent detection performance, effectively addressing juvenile fish feeding behavior detection.

This paper reviews the application and potential of cloud-edge-end collaborative (CEEC) technology in the field of freshwater aquaculture, a rapidly developing sector driven by the growing global demand for aquatic products. The sustainable development of freshwater aquaculture has become a critical challenge due to issues such as water pollution and inefficient resource utilization in traditional farming methods. In response to these challenges, the integration of smart technologies has emerged as a promising solution to improve both efficiency and sustainability. Cloud computing and edge computing, when combined, form the backbone of CEEC technology, offering an innovative approach that can significantly enhance aquaculture practices. By leveraging the strengths of both technologies, CEEC enables efficient data processing through cloud infrastructure and real-time responsiveness via edge computing, making it a compelling solution for modern aquaculture. This review explores the key applications of CEEC in areas such as environmental monitoring, intelligent feeding systems, health management, and product traceability. The ability of CEEC technology to optimize the aquaculture environment, enhance product quality, and boost overall farming efficiency highlights its potential to become a mainstream solution in the industry. Furthermore, the paper discusses the limitations and challenges that need to be addressed in order to fully realize the potential of CEEC in freshwater aquaculture. In conclusion, this paper provides researchers and practitioners with valuable insights into the current state of CEEC technology in aquaculture, offering suggestions for future development and optimization to further enhance its contributions to the sustainable growth of freshwater aquaculture.

Fry feeding in recirculating aquaculture systems (RAS) has gained prominence following China’s ban on fishing in the Yangtze River. Although previous researches have focused on dynamic adjustments to adult fish feeding status, research on fry feeding has been subsequently neglected. To fill this research gap, a precise fry feeding method was developed, comprising four main components: a fry feeding status detection module, a feeding control module, a precise feed discharging module, and a variable feed distribution module. The detection module utilizes the improved FFD-YOLO network which incorporates GhostNet, BiFPN and CA attention to detect fry feeding status, and real-time feeding decisions were made accordingly. Numerical simulations using Python were conducted to calculate the optimal feed coverage ratio, and Fuzzy-PID control was employed to rapidly adjust the rotational speed of the spreading disc. The experiments demonstrate that the FFD-YOLO algorithm achieved a precision of 91.33 %, a recall rate of 74.15 %, and a mAP_0.5 of 85.06 %, with a detection speed of 75 frames per second (FPS). Feeding distribution coverage ratios of 40 % and 80 % were recommended based on simulation results. The experimental results demonstrated that when feeding based on clear images, the errors of discharge and distribution were less than 10.2 % and 12.6 %, respectively. In contrast, when feeding based on blurred images, the errors exceeded 18.4 % and 24.1 %, respectively. Control experiments demonstrated that the proposed method can promote the growth of fry. This study provides a significant reference for future research on automatic fry feeding in industrial RAS.

Pellet feed is widely used in fry feeding, which cannot sink to the bottom in a short time, so most fries eat in shallow underwater areas. Aiming at the characteristics of fry feeding, we present herein a nondestructive and rapid detection method based on a shallow underwater imaging system and deep learning framework to obtain fry feeding status. Towards this end, images of fry feeding in shallow underwater areas and floating uneaten pellets were captured, following which they were processed to reduce noise and enhance data information. Two characteristics were defined to reflect fry feeding behavior, and a YOLOv4-Tiny-ECA network was used to detect them. The experimental results indicate that the network works well, with a detection speed of 108FPS and a model size of 22.7 MB. Compared with other outstanding detection networks, the YOLOv4-Tiny-ECA network is better, faster, and has stronger robustness in conditions of sunny, cloudy, and bubbles. It indicates that the proposed method can provide technical support for intelligent feeding in factory fry breeding with natural light.

针对无人机精确植保过程中,果树冠层区域颜色特征和杂草相似度较高、难以分割等问题,采用基于超像素特征向量的果树冠层分割方法,以消除不同杂草特征对树冠分离的干扰,减小农药喷雾区域,节省农药使用量.通过分析无人机采集合成的样本图像在HSV彩色空间上色调与饱和度的分布情况,选取合适的阈值范围,提取样本图像中包含果树冠层与杂草的绿色区域,将提取的绿色区域RGB图像转换生成Lab和HSV彩色空间模型下的图像,然后运用简单的线性迭代聚类(Simple linear iterative clustering,SLIC)超像素分割算法将RGB图像预设分割成250个超像素单元,结合超像素的分割信息与RGB图像、Lab图像、HSV图像以及灰度图,提取超像素单元的特征向量,随机选取25％的超像素样本的特征向量作为SVM分类器的训练集,利用SVM分类器对所有样本进行预测分类,实现果树冠层与杂草分割.将基于超像素特征向量的方法和基于光谱阈值、K-means聚类的2种方法进行对比分析,结果显示,基于超像素特征向量的方法在识别果树冠层位置方面生产者精度为90.83％,在提取果树冠层轮廓上F测度值为87.62％,总体分割性能优于后两种方法.说明,基于超像素特征向量的方法能够较为准确地分割果树冠层与杂草,为实现无人机在果园中精确植保提供重要支撑.

Soil steam disinfection (SSD) technology is an effective means of eliminating soil borne diseases. Among the soil cultivation conditions of facility agriculture in the Yangtze River Delta region of China, the clay soil particles (SPs) are fine, the soil pores are small, and the texture is relatively viscous. When injection disinfection technology is applied in the clay soil, the diffusion of steam is hindered and the heating efficiency is substantially affected. To increase the heating efficiency of SSD, we first discretized the continuum model of Philip and De Vries into circular particle porous media of different sizes and random distribution. Then with Computational Fluid Dynamics (CFD) numerical simulation technology, a single-injection steam disinfection model for different SP size conditions was constructed. Furthermore, the diffusion pattern of the macro-porous vapor flow and matrix flow and the corresponding temperature field were simulated and analyzed. Finally, a single-pipe injection steam disinfection verification test was performed for different SP sizes. The test results show that for the clay soil in the Yangtze River Delta region of China, the test temperature filed results are consistent with the simulation results when the heat flow reaches H = 20 cm in the vertical direction, the simulation and test result of the heat flow in the maximum horizontal diffusion distance are L = 13 cm and 12 cm, respectively. At the same disinfection time, the simulated soil temperature change trend is consistent with the test results, and the test temperature is lower than the simulated temperature. The difference between the theoretical temperature and the experimental temperature may be attributed to the heat loss in the experimental device. Further, it is necessary to optimize the CFD simulation process and add the disintegration and deformation processes of soil particle size with the change of water content. Furthermore, the soil pores increase as the SP size increases and that a large amount of steam vertically diffuses along the macropores and accumulates on the soil surface, causing ineffective heat loss. Moreover, soil temperature distribution changes from oval (horizontal short radius/vertical long radius = 0.65) to irregular shape. As the SP size decreases, the soil pore flow path becomes fine; the steam primarily diffuses uniformly around the soil in the form of a matrix flow; the diffusion distance in the horizontal direction gradually increases; and the temperature distribution gradually becomes even, which is consistent with the soil temperature field simulation results. Similar to the energy consumption analysis, the maximum energy consumption for SP sizes>27mm and <2mm was 486and 477kJ, respectively. Therefore, proper pore growth was conducive to the diffusion of steam, but excessive pores cause steam to overflow, which increased energy consumption of the system. Considering that the test was carried out in an ideal soil environment, the rotary tiller must be increased for fine rotary tillage in an actual disinfection operation. Although large particles may appear during the rotary tillage process, an appropriate number of large particles contributes to the formation of a large pore flow, under the common effect of matrix flow, it will simultaneously promote greater steam diffusion and heating efficiency. The above theoretical research has practical guiding significance for improving the design and disinfection effect of soil steam sterilizers in the future.

To make canopy information measurements in modern standardized apple orchards, a method for canopy information measurements based on unmanned aerial vehicle (UAV) multimodal information is proposed. Using a modern standardized apple orchard as the study object, a visual imaging system on a quadrotor UAV was used to collect canopy images in the apple orchard, and three-dimensional (3D) point-cloud models and vegetation index images of the orchard were generated with Pix4Dmapper software. A row and column detection method based on grayscale projection in orchard index images (RCGP) is proposed. Morphological information measurements of fruit tree canopies based on 3D point-cloud models are established, and a yield prediction model for fruit trees based on the UAV multimodal information is derived. The results are as follows: (1) When the ground sampling distance (GSD) was 2.13–6.69 cm/px, the accuracy of row detection in the orchard using the RCGP method was 100.00%. (2) With RCGP, the average accuracy of column detection based on grayscale images of the normalized green (NG) index was 98.71–100.00%. The hand-measured values of H, SXOY, and V of the fruit tree canopy were compared with those obtained with the UAV. The results showed that the coefficient of determination R2 was the most significant, which was 0.94, 0.94, and 0.91, respectively, and the relative average deviation (RADavg) was minimal, which was 1.72%, 4.33%, and 7.90%, respectively, when the GSD was 2.13 cm/px. Yield prediction was modeled by the back-propagation artificial neural network prediction model using the color and textural characteristic values of fruit tree vegetation indices and the morphological characteristic values of point-cloud models. The R2 value between the predicted yield values and the measured values was 0.83–0.88, and the RAD value was 8.05–9.76%. These results show that the UAV-based canopy information measurement method in apple orchards proposed in this study can be applied to the remote evaluation of canopy 3D morphological information and can yield information about modern standardized orchards, thereby improving the level of orchard informatization. This method is thus valuable for the production management of modern standardized orchards.

Perception of the fruit tree canopy is a vital technology for the intelligent control of a modern standardized orchard. Due to the complex three-dimensional (3D) structure of the fruit tree canopy, morphological parameters extracted from two-dimensional (2D) or single-perspective 3D images are not comprehensive enough. Three-dimensional information from different perspectives must be combined in order to perceive the canopy information efficiently and accurately in complex orchard field environment. The algorithms used for the registration and fusion of data from different perspectives and the subsequent extraction of fruit tree canopy related parameters are the keys to the problem. This study proposed a 3D morphological measurement method for a fruit tree canopy based on Kinect sensor self-calibration, including 3D point cloud generation, point cloud registration and canopy information extraction of apple tree canopy. Using 32 apple trees (Yanfu 3 variety) morphological parameters of the height (H), maximum canopy width (W) and canopy thickness (D) were calculated. The accuracy and applicability of this method for extraction of morphological parameters were statistically analyzed. The results showed that, on both sides of the fruit trees, the average relative error (ARE) values of the morphological parameters including the fruit tree height (H), maximum tree width (W) and canopy thickness (D) between the calculated values and measured values were 3.8%, 12.7% and 5.0%, respectively, under the V1 mode; the ARE values under the V2 mode were 3.3%, 9.5% and 4.9%, respectively; and the ARE values under the V1 and V2 merged mode were 2.5%, 3.6% and 3.2%, respectively. The measurement accuracy of the tree width (W) under the double visual angle mode had a significant advantage over that under the single visual angle mode. The 3D point cloud reconstruction method based on Kinect self-calibration proposed in this study has high precision and stable performance, and the auxiliary calibration objects are readily portable and easy to install. It can be applied to different experimental scenes to extract 3D information of fruit tree canopies and has important implications to achieve the intelligent control of standardized orchards.