基于改进的PSO进化神经计算进行苹果颜色快速分级

为克服在苹果颜色分级中存在的速度慢、误差大等缺点,基于再现群智能的粒子群进化算法和神经计算技术,提出了一种新颖、快速的智能分级方法,即首先通过计算机视觉技术获取苹果表面颜色的色度,并提取其特征;然后采用改进的带自适应惯性权值的粒子群优化算法训练神经网络结构,最后用训练好的神经网络进行苹果颜色分级。实际应用表明该方法切实可行且效果显著,不仅分级速度快,而且分级正确率高达98%以上。     

油梨嫁接苗分级标准探讨

通过逐步聚类分析方法,初步探讨了油梨嫁接苗的分级标准。分析得出以苗高和茎粗为油梨嫁接苗的分级质量指标,将油梨嫁接苗分为三级,即Ⅰ级苗：苗高≥74cm,茎粗≥1.13cm;Ⅱ级苗：74cm＞苗高≥66 cm,1.13cm＞茎粗≥0.93cm;Ⅲ级苗：苗高＜66cm,茎粗＜0.93cm。     

Limnology of Macrobrachium amazonicum grow‐out ponds subject to high inflow of nutrient‐rich water and different stocking and harvest management

We evaluated the water characteristics and particle sedimentation in Macrobrachium amazonicum (Heller 1862) grow‐out ponds supplied with a high inflow of nutrient‐rich water. Prawns were subject to different stocking and harvesting strategies: upper‐graded juveniles, lower‐graded juveniles, non‐graded juveniles+selective harvesting and traditional farming (non‐grading juveniles and total harvest only). Dissolved oxygen, afternoon N‐ammonia and N‐nitrate and soluble orthophosphate were lower in the ponds in comparison with inflow water through the rearing cycle. Ponds stocked with the upper population fraction of graded prawns showed higher turbidity, total suspended solids and total Kjeldahl nitrogen than the remaining treatments. An increase in the chemical oxygen demand:biochemical oxygen demand ratio from inlet (4.9) to pond (7.1–8.0) waters indicated a non‐readily biodegradable fraction enhancement in ponds. The sedimentation mean rate ranged from 0.08 to 0.16 mm day^?1 and sediment contained >80% of organic matter. The major factors affecting pond ecosystem dynamic were the organic load (due to primary production and feed addition) and bioturbation caused by stocking larger animals. Data suggest that M. amazonicum grow‐out in ponds subjected to a high inflow of nutrient‐rich water produce changes in the water properties, huge accumulation of organic sediment at the pond bottom and non‐readily biodegradable material in the water column. However, the water quality remains suitable for aquaculture purposes. Therefore, nutrient‐rich waters, when available, may represent a source of unpaid nutrients, which may be incorporated into economically valued biomass if managed properly.     

Intensive rearing of juvenile crayfish (Pacifastacus leniusculus,Astacidae) during the first 6 months: effects of size grading

Recent advances in crayfish feeding have enabled the development of size grading studies from the start of first‐feeding. A 180‐day experiment aimed at intensive rearing of Pacifastacus leniusculus was carried out under controlled conditions, evaluating the effects of size grading at two different periods from the onset of exogenous feeding. Stage 2 juveniles were stocked in fibreglass tanks at a density of 100 m², and fed a dry diet for salmonids combined with restricted amounts of Artemia cysts. Five groups were tested: no grading, grading at 60 days (large and small size) and grading at 100 days (large and small size). After 6 months, no significant differences were found in the survival among groups (mean: 73.06%). The highest final growth (pooled results from upper and lower classes: 17.39 mm carapace length, 1.43 g weight) was achieved by the crayfish sorted at 60 days, showing significant differences from the ungraded group. Smaller crayfish graded at 60 days grew significantly faster than smaller crayfish graded at 100 days. The food conversion ratio was lower in the graded groups (mean: 2.64), showing significant differences from the ungraded group (3.23). This study shows that size grading allows a better performance and an improved feeding efficiency.     

初烤烟叶专业化分级效率研究

设计4种工位的流程化分级方式,分别对第2、第4、第6炕次烟叶实施专业分级试验,研究不同的流程化分级方式对烟叶分级速度与质量的影响。结果表明:二工位和三工位流程化分级速度最快,分别为60.75、60.51kg/(人.工),较一工位分级分别多6.82、6.58 kg/(人.工),较四工位分级分别多9.38、9.14 kg/(人.工);四工位、三工位流程化分级时各炕次烟叶分级质量最好,烟叶等级合格率分别为91.78%、90.38%,比一工位分组分别高24.23及22.83百分点,比二工位分级分别高14.55及13.15百分点。对比4种不同方式的烟叶分级速度及分级质量,以三工位流程化分级效率最高。     

Damage of potato tubers,a review

Summary Damage to potato tubers due to mechanical forces is among the most important causes of loss of quality reported throughout the world. The causes of such damage are mechanical injuries that occur primarily in association with harvesting and grading. The extent of damage is dependent on the condition of the tubers and also reflects the history of the potatoes. Although important advances have been made, there is still a considerable need for research on this subject since knowledge of the combined effects of essential physiological proceses within the cells is still lacking. The avoidance of exposure to mechanical forces constitutes the most effective protection against tuber damage.     

基于加权随机森林的番茄氮元素缺乏分级模型研究

基于叶面颜色特征建立番茄氮元素缺乏分级模型判别准确率可达08以上。夏季定植的番茄叶片表面会覆盖粘质腺毛,粘质腺毛利于番茄吸收水分和营养元素,相同营养液氮离子浓度下叶片黄化过程异于未覆盖粘质腺毛的叶片。故仅基于叶面颜色特征建立分级模型,其准确率降至0.65。覆盖粘质腺毛番茄其叶片周长和叶面积两个形状特征均小于未覆盖粘质腺毛的番茄叶片,本文将番茄叶片两个形状特征结合原有叶面颜色特征共同作为模型输入,建立新的番茄氮元素缺乏分级模型。搭建图像采集系统,该图像采集单元由树莓派和其相机模块构建,使用WiFi或4G网络完成智能手机、图像采集单元、本地计算机之间无线数据传输。智能手机通过Web界面可远程控制采集图像并将图像传输到云平台存储。本地计算机对图像进行预处理提取叶片形状、颜色特征后输入模型进行预测,并输出预测结果。试验结果表明,图像采集系统春季和夏季平均温度在19.7~28.3℃范围内,光照在1.125~9.543lx范围内均可正常使用,采集的图像经预处理分割后降低了环境光线的影响。使用优化后的加权随机森林模型,基于形状特征和颜色特征相结合的叶片氮元素缺乏分级判别准确率可达0.83。     

基于机器视觉的红掌检测分级方法

中国盆花销售量随着消费水平的提高而迅速攀升,四大盆花之一的红掌在规模化生产中存在分级标准掌控不统一、分级结果不稳定以及效率低下等问题,制约了红掌产品的规模化生产,影响了红掌的出品质量。该研究提出基于机器视觉技术的红掌分级检测方法,对90株红掌的水平和竖直方向图像采集,经过二值化等处理,从侧视轮廓、俯视轮廓及佛焰苞轮廓特征三方面获取红掌植株的高度、冠幅、佛焰苞片数、苞片横径4项指标的信息, 针对每一项指标提出检测方法。试验结果表明,该方法对植株高度测量误差小于5.4 mm;针对植株冠幅提出了当量直径测量方法,佛焰苞片数测量正确率可达98.9%,苞片横径相对误差最大为6.52%。针对试验所选90盆红掌分级成功率达到97.8%。研究表明,利用机器视觉技术能够很好地实现对红掌盆花的在线检测分级。