Abstract:The poultry breeding in China is moving towards precision breeding, automatic breeding, green breeding, etc. Collaborative research, development and application of industrial poultry farming technology and information technology are the keys to promote the healthy and sustainable development of poultry farming industry, and are of great significance to improve the scale, standardization and intelligent level of poultry breeding, improve the overall output and economic benefits, and promote the transformation and upgrading of modern poultry breeding industry. Focusing on the three technical fields of intelligent equipment, intelligent algorithm, management and control platform for poultry breeding, this paper analyzed the latest research, application progress and existing problem of intelligent breeding house, environmental monitoring and control, intelligent feeding, epidemic prevention, patrol inspection and harmless feces disposal equipment, poultry behavior detection, inventory technology, weight estimation and health status evaluation algorithm, and factory poultry breeding management and control platform. The weak links and development trend of poultry breeding technology were pointed out, and suggestions for the development and improvement of factory poultry breeding were put forward, providing references for the green high-quality transformation and upgrading of modern poultry breeding industry as well as healthy and sustainable development in China.

Abstract:Factory pig farming is an important part of the modernization of animal husbandry and the inevitable trend of the development of pig industry in China. The pig industry in China is facing the outstanding problems of low productivity level, low health management level, low utilization rate of intelligent equipment and high breeding cost. In this paper, we analyzed the research and development status of technology and equipment in the construction of intelligent pig factory from four aspects: Pig welfare and healthy breeding technology, pig house air purification technology, pig growth and health perception technology, pig precision feeding and breeding robot, and looked forward to the key direction of intelligent pig factory construction, so as to provide a reference for the creation of intelligent pig factory in China.

Abstract:智能识别技术是水产养殖由粗放型向集约型转变的关键技术。水产养殖中的智能识别是通过研究并利用机器视觉和机器学习技术实现水下生物和环境的监测，并对生产管理中出现的问题进行判断、分析和预测，以实现自动化养殖为目的。本文从生物的物种识别与分类、年龄识别、性别识别和行为识别4个方面分析了水产养殖中智能识别技术的研究和发展现状，阐述了水产养殖中采用的主要智能识别技术和原理，并对今后水产养殖中智能识别技术的发展进行了展望，以期为中国渔业现代化、智慧化发展提供参考和新思路。

Abstract:Weed control is an important issue that must be faced in agricultural production. With the integration of robotics and automation technology into agricultural production, various weeding robots emerge as the times require, effectively reducing the pollution of chemicals to the environment. In this paper, the research status of intelligent sensing technology, robot platform and weeding device of weeding robot are reviewed, and the shortcomings of crop row and weed identification technology, structure of weeding robot platform and intelligent control method of mechanical weeding device are analyzed. The future development trend of intelligent weeding robot is prospected from four aspects of intelligent perception, precise weeding, efficient operation and intelligent management.

Abstract:Using field monitoring technology to collect crop information, we can obtain the growth of field crops in real time and make corresponding decisions, which is important for improving the yield and quality of crops. The rapid monitoring, information acquisition and analysis of field crops have become a hot topic of research today because traditional crop field monitoring methods rely on manual sampling and measurement, which have some shortcomings of low efficiency, strong subjectivity and single characteristic. This paper analyzed the current research status of field crop monitoring technology at home and abroad in terms of three aspects of acquisition targets, monitoring platforms and different data (information) analysis methods, summarized the current problems of field crop monitoring in China. Finally, some suggestions of the future development were put forward in terms of monitoring technology innovation, information analysis technology, data (information) standardization and sharing, infrastructure and extension, with the aim of providing a reference for innovation and industrialization of field crop monitoring technology in China.

Abstract:Objective In order to design an intelligent feeding Internet of Things system for lactating sows，and realize the remote monitoring of feeding status of lactating sows.Method The system transmited the custom TCP communication protocol through Netty to realize the function of data transmission and instruction reply with the terminal equipment. The design of human-computer interaction interface was carried out by using SpringBoot and Vue front-end and back-end separation architecture, including pig farm production detail interface, sow feeding information query interface and statistical data download interface.Result The test results showed that the average response time of the system was 0.33 s with 3000 connections, and the amount of data processed per unit time ranged from 750 to 1180. After the system was added into the custom business thread pool, the number of data processed per unit time increased by 250, and the processing capacity increased by 31%.Conclusion The system meets the practical application requirements of connection management and data processing for intelligent feeding equipment in nursing house.

Abstract:Objective There are problems in the course of river crab farming due to water level changes as well as slow convergence and low accuracy of the path planning algorithm of unmanned craft. Therefore, a multi-objective particle swarm-ant colony fusion algorithm for unmanned vehicle path planning was presented to improve the adaptability and optimization ability of the algorithm.Method Firstly, the factors such as crab pond environment and breeding law were analyzed, and the environmental model of static water depth in grid was established. Secondly, to cope with the issues of inadequate local point feeding and sub-optimal paths in coverage traversal baiting, a modified particle swarm optimization (PSO) algorithm based on multi-objective was presented by non-linear adjustment of inertia parameters and learning factors. The initial pheromone of the ant colony algorithm was adjusted, and the pheromone volatility factor and heuristic expectation function of the ant colony algorithm were improved to present an adaptive ant colony optimization (ACO) algorithm. Finally, to address the shortcomings of a single algorithm for finding the best, a fusion of PSO-ACO was utilized to realize multi-objective global path planning for baiting vessels.Result The simulation results showed that the PSO-ACO algorithm not only had good environmental adaptability but also improved the efficiency and accuracy of multi-target path finding under different environmental baiting strategies. The PSO-ACO algorithm saved the running time by 32%, shortened the path distance by 9.78%, reduced the number of iterations by 62.88% and reduced the number of inflection points by 44.45%.Conclusion The proposed multi-objective path planning algorithm is suitable for crab pond culture with variable environment, and has good application value.

Abstract:目的 为解决传统水稻考种机谷粒表型分析算法在功能和效率上的局限性，针对穗上谷粒原位计数和被遮挡谷粒几何特征还原设计一种基于深度学习的轻量级通用算法框架。方法 将穗上谷粒原位计数与被遮挡谷粒还原这2个复杂任务分别拆解为2个阶段，将其核心阶段建模为I2I问题。基于MobileNet V3设计1种能够解决I2I问题的轻量级网络架构，并针对2个任务的特点分别设计了数据集图像制作方法，选择合适的优化策略和超参数对其进行训练。训练结束后，使用TensorFlow Lite runtime解释器将模型部署在考种机的树莓派4B开发板上，并进行测试。结果 该算法在穗上谷粒计数任务中具有良好的准确性、快速性，且具有一定的泛化性能。在被遮挡谷粒的形状还原任务中，该算法所还原的谷粒图像在面积、周长、长度、宽度和颜色分数评价指标中准确率均达到97%以上。结论 该算法能够有效地完成穗上谷粒计数和被遮挡谷粒的还原任务，且具有轻量级的优点。

Abstract:Objective To accurately predict the water and temperature of the arable layer using the correlation between soil near surface air temperature and humidity and soil internal parameters, and serve for the realization of fine agricultural planting management. Method Aiming at the actual needs of soil tillage layer moisture and temperature prediction in training set acquisition and model verification, an internet of things data acquisition system based on embedded system and narrow band internet of things (NB-IoT) wireless communication technology was designed. A model combination strategy was explored based on the deep Q network (DQN) deep reinforcement learning algorithm. Based on the weighted combination of long short-term memory (LSTM), gated recurrent unit (GRU) and Bi-directional long-short term memory (Bi-LSTM), the DQN-L-G-B combination prediction model was obtained. Result The data acquisition system achieved long-term stable and reliable collection of time series environmental data with equal intervals, and provided accurate training set and verification set data for soil moisture, temperature time series prediction based on deep learning. Compared with models such as LSTM, Bi-LSTM, GRU and L-G-B, the DQN-L-G-B combined model not only lowered the root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) in the prediction of moisture and temperature on the tillage layer of the two soil types (loam and sand), but also increased R² by about 0.1%. Conclusion Through the internet of things data acquisition system and the DQN-L-G-B combined model, the accurate prediction of soil moisture and temperature in the cultivated layer based on soil near surface air temperature and humidity can be effectively completed.

Abstract:Objective Develop a method to improve the potato (Solanum tuberosum) leaf area index (LAI) estimation accuracy using the UAV multiple spectral wavebands and texture information.Method The DJI P4M drone was used to collect multispectral images of the southern winter potato at seedling period, budding period and tuber swelling period from February to April 2021. LAI data were measured by LAI-2000 canopy analyzer. The spectral and texture characteristics of images were extracted. The correlations between vegetation index, texture characteristics and LAI were analyzed. The selected characteristic variables were analyzed based on subset of adjusted R²_adj. The principal component analysis was used to fuse spectrum and texture features, and the principal component analysis-multiple linear regression (PCA-MLR) model was used to estimate potato LAI.Result From the seedling period to the tuber swelling period, the PCA-MLR estimation model was better than texture multiple linear regression (T-MLR) and vegetation index multiple linear regression (VI-MLR) model, with R² of 0.73, 0.59 and 0.66 respectively. Conclusion This study proposed a method of PCA-MLR to estimate the potato LAI and improve the levels of the potato growth monitoring and field management.

Abstract:Objective In view of the problems that the existing experimental equipment for root phenotype detection is expensive, requires special personnel to operate, and cannot perform rapid in situ non-destructive detection of root phenotypes, this paper proposes a non-destructive detection method for radish root phenotype based on electrical resistance tomography (ERT) and deep residual network (ResNet). Method Firstly, the ERT positive problem of the radish-agar field for different cases was performed by COMSOL, and a large amount of boundary voltage data was obtained. Secondly, the nonlinear mapping relationship between the internal conductivity distribution and the boundary voltage inside the radish-agar field was modeled based on ResNet, and image of radish-agar field were reconstructed. Finally, a set of radish root phenotype detection device was developed based on ERT, and the experimental verification was carried out.Result The radish root phenotype detection method based on ERT and ResNet could achieve sustainable nondestructive detection of radish root phenotype, and the experimental setup was simple to operate and low cost, while the relative error of image reconstruction was less than 5%.Conclusion The method of radish root phenotype detection based on ERT can achieve nondestructive detection of radish root phenotype with high imaging accuracy combining with ResNet algorithm. This method can be effectively used for the detection of radish root phenotypes.

Abstract:目的 针对实际生产场景中番茄苗期生长遇到的高温胁迫问题，提出一种基于热红外和RGB图像的番茄苗期高温胁迫检测方法。方法 首先，通过番茄苗期热红外图像反演获取番茄冠层温度参数，采用偏最小二乘(Partial least squares, PLS)模型提取冠层温度特征指标；然后，建立采用3种不同主干特征提取网络的Mask-RCNN模型，通过迁移学习的方式将番茄苗期RGB图像输入Mask-RCNN模型，进行高温胁迫症状实例分割，得到番茄苗期胁迫症状特征指标；最后，利用提取的温度和胁迫症状特征指标构建分级数据集，输入高温胁迫分级模型，得到高温胁迫等级。结果 基于PLS模型提取的冠层温度特征指标累计贡献率达95.45%；基于ResNet101+Mask-RCNN的高温胁迫症状分割网络对番茄苗期轻度和重度胁迫的分割精度最高，均值平均查准率(Mean average precision, mAP)分别为77.3%和73.8%；基于温度和胁迫症状特征指标构建的4种高温胁迫分级模型中，反向传播神经网络(Back propagation neural network, BPNN)获得最好的高温胁迫分级效果，分级准确率达95.6%。结论 该方法对番茄苗期高温胁迫检测效果较好，可为番茄苗期高温胁迫早期精准检测和快速自动预警提供技术支撑。

Abstract:Objective In order to provide decision-making basis for subsequent litchi flower thinning, fruit retention and precise fertilization application, this work evaluated the flowering rate of each litchi by analyzing UAV remote sensing images of litchi canopy.Method The remote sensing image of each litchi canopy was segmented through instance segmentation algorithm. The flowering rates were classified into four categories combining with comprehensive judgment of horticultural experts, which were 0, 10%?20%, 50%?60%, 80% and above. ResNet50, ResNeXt50 and ShuffleNetv2 were adopted to compare flowering rate classification. Due to the great advantages in recognition accuracy, number of parameters, training and verification time, ShuffleNetv2 was adopted as the instance segmentation algorithm, and optimized by introducing the spatial attention module (SAM) to increase the model’s learning of location information, and improve the accuracy of litchi canopy flowering classification without significantly increasing the number of parameters. Result Through comparison of the mainstream deep learning algorithms, the classification accuracy of ResNet50, ResNeXt50 and ShuffleNetv2 reached 85.96%, 87.01% and 86.84% respectively, and the improved ShuffleNetv2 reached 88.60%, higher than the above three algorithms. The verification time of single canopy image on test set using ResNet50, ResNeXt50, ShuffleNetv2 and the improved ShuffleNetv2 were 8.802, 9.116, 7.529 and 7.507 ms respectively, showing that the improved ShuffleNetv2 single canopy image got the shortest verification time. Conclusion The improved ShuffleNetv2 can excavate and learn more detailed flowering information of litchi canopy, with high recognition accuracy and great advantages in the evaluation of litchi flowering, providing an intelligent decision support for flower protection and sparseness, and precise control of production.

Abstract:Objective The semantic segmentation technology was used to automatically identify mango and its skin defects, to realize the quality evaluation and sorting of mango and provide a reference for the rapid and nondestructive testing of mango quality.Method Mango skin defect images in multi-scene of natural environment were collected for model training and testing. Atrous spatial pyramid pooling (ASPP) in DeepLabV3+ was replaced by joint pyramid upsampling (JPU) structure, and Xception model in DeepLabV3+ was replaced by Atrous-ResNet model. Class pixel accuracy (CPA), mean pixel accuracy (MPA) and mean intersection over union (MIoU) were used as the accuracy evaluation indexes of each model. Result JPU module was used to replace ASPP module, and Atrous convolution was applied to ResNet network which was conductive to increase the receptive field of the model. In general, the predicted boundary was smoother, and the identification of small defects was more accurate. The comparison with SegNet and LinkNet algorithms showed that Atrous-ResNet model had higher accuracy, with CPA slightly improved, MPA was up 3.79 percent point and MIoU was up 4.57 percent. Atrous-ResNet model had better identification effects. Conclusion The method based on semantic segmentation is feasible for mango skin defect recognition. Compared with SegNet and LinkNet algorithms, Atrous-ResNet model has higher recognition accuracy.

Abstract:Objective In order to solve the problem of low recognition accuracy due to similar color of the background and new shoots, to realize the automatic monitoring of citrus shoot stage and explore the improved method of algorithm, the machine vision technology was used to carry out the research on intelligent perceiving the growth stage of citrus shoot.Method According to the characteristics of features extracted from different convolutional layers and the role of different attention mechanism, an improved YOLOX-Nano intelligent recognition model based on multi-attention mechanism was proposed, and a diversified orchard dataset for pre-training was established.Result The improved YOLOX-Nano algorithm achieved the mAP (Mean average precision) of 88.07% using the orchard dataset as a pre-training dataset. Compared with the model of YOLOV4-Lite series, the improved model significantly improved the recognition accuracy with less parameters and calculation. Compared with YOLOV4-MobileNetV3 and YOLOV4-GhostNet, the mAP of improved model increased by 6.58% and 6.03% respectively.Conclusion The improved model has greater advantage for lightweight deployment at orchard monitoring terminal. The findings provide feasible data and technical solutions for agricultural real-time perception and intelligent monitoring.

Abstract:目的 为实现复杂背景下广佛手发病早期的病虫害快速精准识别，提出一种基于YOLOv5-C的广佛手病虫害识别方法。方法 使用YOLOv5s网络模型作为基础网络，通过引入所提出的多尺度特征融合模块，提高网络模型的特征提取与特征融合能力，均衡提高每一类广佛手病虫害的识别准确率；使用注意力机制模块提高网络模型对病虫害目标特征信息的关注度，弱化复杂背景的干扰信息，提高网络模型的识别准确率；利用改进的C3-SC模块替换PANet结构中的C3模块，在不影响网络模型识别性能的条件下减少网络模型的参数。结果 基于YOLOv5-C的复杂背景下的广佛手病虫害识别，F1分数为90.95%，平均精度均值为93.06%，网络模型大小为14.1 Mb，在GPU上每张图像平均检测时间为0.01 s。与基础网络YOLOv5s相比，平均精度均值提高了2.45个百分点，7个类别识别的平均准确率的标准差由7.14减少为3.13，变异系数由7.88%减少为3.36%。平均精度均值比RetinaNet、SSD、Efficientdet和YOLOv4模型分别高22.30、20.65、4.84和2.36个百分点。结论 该方法能快速准确地识别复杂背景下广佛手病虫害目标，可为广佛手种植产业的智能化管理提供参考。

Abstract:Objective In order to extract the working path in the agricultural robot navigation system, we proposed an algorithm for identifying the drivable area between rows of fruit trees with the sky as the background in a complex environment.Method The tree crown and the background sky were separated by the blue component (B component), and the Otsu algorithm was improved to achieve a better effect of segmentation. After morphological processing, according to the regularity of tree top distribution, dynamic threshold was used to find “V-shaped” region of interest and extract feature points. After the interference points were eliminated by Theil-Sen robustness regression, the straight line at the tree top was fitted by random sample consensus (RANSAC) algorithm, the slope of the straight line at the edge of the drivable area was obtained through the slope transformation relationship, and the key point coordinates were obtained using the information of the feature points after elimination and the threshold elimination. Taking the slope as the constraint condition, the linear equation of the edge of the drivable area was obtained by substituting the key points. The least square method was used to fit the data for realizing the recognition of the drivable area.Result The experimental results showed that compared with Theil-Sen algorithm and RANSAC algorithm, the average deviation angle of the double robustness regression algorithm in this paper was reduced by 8.28% and 9.88%, the standard deviation was reduced by 6.25% and 22.89%, and the accuracy was improved by 4.64% and 10.49%. Conclusion The research results can provide research ideas for the drivable area recognition and path extraction of agricultural robots in the complex environment of most standardized orchards.

Abstract:Objective To determine the parameters of discrete element simulation model in the mechanized processing of fresh lotus seeds, and provide data references for the mechanized processing simulation test of fresh lotus seeds.Method The calibration of discrete element simulation parameters of fresh lotus seeds was carried out by EDEM simulation software. The accumulation angle and repose angle of the actual fresh lotus seeds were measured by seed drop test with ‘Space lotus 36’ from Honghu, Hubei. Based on the Hertz-Mindlin (no slip) contact model, a simulation test of fresh lotus seed drop was conducted, and the error between the measured and simulated values of fresh lotus seed accumulation angle and repose angle was used as the test index to determine the contact parameters with significant effects on accumulation angle and repose angle through the Plackett-Burman test. The steepest climb test was conducted to determine the optimal contact parameter combinations in discrete element model for fresh lotus seeds. The actual seed drop verification test was carried out using the hopper with seed drop rate as the test index. The seed drop rates in actual and simulated seed drop verification tests were compared to verify the reliability of the optimal parameter combination.Result The effects of static friction coefficient between lotus seeds and rolling friction coefficient between lotus seeds on the accumulation angle were highly significant (P<0.01); The effects of rolling friction coefficient between lotus seeds on the repose angle were highly significant (P<0.01), and the effects of static friction coefficient between lotus seeds and static friction coefficient between lotus seeds and plexiglass on the repose angle were significant (P<0.05). The optimal combination of contact parameters was 0.4 for static friction coefficient between lotus seeds, 0.02 for rolling friction coefficient between lotus seeds, and 0.4 for static friction coefficient between lotus seeds and plexiglass. The results of the drop verification test showed that the maximum relative error of fresh lotus seed drop rate between the actual test and the simulation test was not exceeding 3.65%.Conclusion The contact parameters of the calibrated discrete element simulation model for fresh lotus seeds are accurate and reliable, and the findings can provide data references for the structural design optimization of lotus seed processing machinery.