基于YOLO v5与短时跟踪的鸡只呼吸道疾病早期检测

针对鸡只呼吸困难这一早期呼吸道疾病显著症状难以检测的问题,提出一种基于YOLO v5与短时跟踪的鸡只呼吸道疾病早期检测方法。对YOLO v5算法进行锚框自适应设置与CIoU Loss (Complete IoU Loss)应用等特定优化后,用于群鸡复杂环境中准确识别鸡头目标并检测是否为张口状态。根据鸡头坐标框交并比实现鸡头目标短时跟踪并获取不同鸡头的短时动作序列,再对动作序列进行分析,判断张口-闭口组合出现的频率,动态检测是否存在鸡只呼吸困难情况。实验结果表明,改进YOLO v5算法检测鸡头目标的mAP为80.1%,张口检测准确率为67.3%,闭口检测准确率为92.8%,基于时间序列的呼吸困难行为检测方法的识别准确率为91.8%,召回率为75%,精准率为67.9%,可为群鸡养殖环境中的鸡只早期呼吸道疾病检测提供参考。     

基于性能指标的农用车辆路径跟踪控制器设计

该文以农用车辆为控制对象,设计了路径跟踪PID控制器。基于性能指标—ISE、IAE、ITAE和ITSE分别整定了PID控制器参数,给出了最优PID控制器,同时进行了仿真对比分析。仿真结果表明,以超调量和调节时间这2个时域指标为评价标准,基于上述4种性能指标准则,系统在所给控制器作用下均能获得令人满意的动态和稳态性能。ISE准则整定的PID控制器使闭环系统单位阶跃响应的超调量为25.55%,调节时间为5.07s。相比ISE准则,由IAE、ITAE和ITSE准则整定的控制器使闭环系统单位阶跃响应的超调量更小,为10.03%,调节时间更短,为3.95s。利用本文方法能够获得较好的PID控制器参数,可为农用车辆控制器设计提供理论依据。     

槽式太阳能温差发电装置跟踪控制系统设计

为提高太阳能发电装置的太阳能利用率及输出功率,以槽式太阳能温差发电装置为研究对象,设计一种以PLC为控制器,MCGS触摸屏为组态界面的跟踪控制系统。该系统通过光照度传感器反馈信息,步进电机控制聚光镜,使得聚光镜绕南北东西两侧旋转跟踪太阳的高度角和方位角,以此提高太阳能发电装置的光电转换效率。为验证槽式太阳能温差发电装置跟踪控制系统的可行性,对装置的发电量进行测量,试验结果表明：装置连续10天给日光温室负载供电,总发电时间为52.01h,共发电2.74kW·h,可满足日光温室育苗的日常用电。装置结构灵活,可通过增加串并混合连接温差发电片的数目,进一步提高该装置的发电量。     

Movements of fur seals following relocation from fish farms

• 1. The movement patterns of Australian and New Zealand fur seals trapped on salmon farms in south‐east Tasmania and relocated hundreds of kilometres away, were monitored using satellite telemetry. Australian fur seals released 470 km away by sea returned to farms after an average of 8.5±4.4 days (n=9 trips) and those released at 140 km returned after 3.2±0.8 days (n=4 trips). New Zealand fur seals (n=5 trips) averaged a return time of 6.8±1.2 days from 300 km.
• 2. When in south‐east Tasmania, both seal species undertook short trips to sea (mean 2.6 days) from haul‐out sites (minimum distance from farms 21 km) with 33% (1.8 days) of this time spent within 5 km of farms. Mean haul‐out duration was 1.1 days.
• 3. In summer, Australian fur seals repeatedly travelled between northern Bass Strait islands and southern Tasmanian waters following the Tasmanian east coast. Seals did not visit farms during this time.
• 4. Southern Tasmanian waters are important foraging grounds for fur seals and potential exists for a substantial number of seals to visit nearby fish farms. Farms provide predictable food resources from penned and escaped salmon, and wild fish attracted to the area.
• 5. Australian fur seals trapped at Tasmanian salmon farms regularly visited breeding colonies on islands in Victoria.

Copyright © 2008 John Wiley & Sons, Ltd.     

Basin-scale advection and population persistence of Calanus finmarchicus

Advection of Calanus finmarchicus in the eastern North Atlantic was analysed using a particle-tracking model based on the Hamburg Shelf Ocean Model (HAMSOM). Quasi-static seasonal mean flowfields were simulated, archived and interpolated to represent a climatological-mean annual cycle. Particles had a simple prescribed depth profile comprising deep overwintering, spring ascent, a shallow-water phase followed by descent to overwintering depth. Export routes for C. finmarchicus from the model area were identified to the south of Greenland and to the north of the Lofoten Basin. Self-sustaining overwintering areas were identified by observing how closely particles returned to their origins after one calendar year. Several such areas were found, notably in the Norway and Lofoten Basins, and in the Færoe–Shetland Channel. The particle tracking was run for up to 10 years to demonstrate persistence of these cycles. Known features of the winter and summer distributions of C. finmarchicus were reproduced by the model. The success of the HAMSOM in simulating both the shallow and deep circulation of the eastern North Atlantic and Norwegian Sea was critical to the identification of these spatio-temporal cycles of C. finmarchicus.     

Multiple spawning events promote increased larval dispersal of a predatory fish in a western boundary current

Transport of larvae by ocean currents is an important dispersal mechanism for many species. The timing and location of spawning can have a large influence on settlement location. Shifts in the known spawning habitat of fish, whether due to climate or the discovery of new spawning stock, can influence the distribution of juveniles and our understanding of connectivity. The globally distributed species; Pomatomus saltatrix, is one such example where a previously unrecognised summer spawning event and a more southern latitudinal extent was recently reported for the southwest Pacific population. Although restrictions are in place to protect the traditional spawning event, the importance of the newly recognised summer spawning event is uncertain. Here, we investigate larval dispersal of P. saltatrix using particle tracking simulations to identify the contributions of the different spawning events to settlement. By modelling dispersal of larvae released in northern and mid‐latitude regions over the Austral spring and summer, we show that the newly recognised mid‐latitude summer spawning event contributes over 50% of the larvae reaching southern latitudes. This is due to a reduced (1–2 days) pelagic larval duration (associated with temperature), resulting in reduced larval mortality, and the seasonal (summer) strengthening of the East Australian Current (EAC) transporting particles ~50 km further south. These findings demonstrate that in dynamic boundary current systems such as the EAC, the final settlement location of larvae that are transported by ocean currents can vary considerably depending on the timing and location of spawning and that multiple spawning events are important for maximum dispersal.     

基于ICR的履带车辆路径跟踪与转向控制算法研究

针对丘陵山区单边制动农用履带车辆路径跟踪精度低、控制次数多、转向偏差大等问题,本文开展不同负载条件下履带车辆路径跟踪控制研究。首先,对履带车辆的转向运动学进行理论分析,并建立履带车辆运动学模型;其次,根据履带车辆单边制动转向特性,提出一种基于瞬时旋转中心（Instantaneous center of rotation,ICR）的大角度转向控制算法,该算法能够根据规划路径的转向点位置与履带车辆转向瞬心,规划出最优的转向目标点,并控制履带车辆在该转向目标点一次性转向到所需航向,与此同时,完成转向控制器设计;最后,开展履带车辆在3种不同负载条件下的仿真试验与田间试验。仿真结果表明,大角度转向控制算法产生的跟踪路径平均误差面积与平均转向控制次数分别降低68.95%、68.77%;田间试验结果表明,大角度转向控制算法产生的跟踪路径平均横向偏差均值、平均转向控制次数与转向点处平均最小偏差分别减少57.27%、33.93%、62.29%,且路径跟踪效果更优,验证了大角度转向控制算法的有效性。试验结果满足履带车辆路径跟踪的要求,为实现农用履带车辆的路径跟踪提供理论基础与参考。     

基于变前视距离的四轮同步转向农机改进纯跟踪控制

路径跟踪控制是提高自主导航系统控制精度的关键。针对在复杂农田作业环境下转弯时纯跟踪算法跟踪精度不高的问题,本文提出了一种基于改进纯追踪模型的四轮同步转向农机路径跟踪控制算法。建立了基于四轮同步转向农机的运动学模型和纯跟踪模型,在此基础上考虑航向误差得到改进纯跟踪模型,进行RTK定位坐标修正,根据量化误差的评价函数搜索前视区域最优目标点,得到最优前视距离。本文算法能实时确定四轮同步转向农机改进纯跟踪模型中的前视距离,使航向误差和横向误差最小化,实现目标点的自适应优化。仿真结果表明,本文方法转弯时平均绝对横向误差减至0.035m,平均绝对航向误差减至0.212°;水田实验结果表明,当四轮同步转向农机作业速度为3.6km/h时,四轮转向农机轨迹跟踪平均绝对横向误差减至0.109m,平均绝对航向误差减至2.799°,转弯跟踪精度显著提高。     

基于亚像素定位的图像边缘检测策略研究

针对图像处理与计算机视觉技术中低对比度、边缘模糊图像的边缘检测问题,参考局部极值与梯度方向两种因素,并结合图像边缘方向趋势,提出了一种单像素边缘跟踪策略。相较于应用广泛的Canny算法,该跟踪策略无需设置全局阈值,实现方式更为简洁、高效;提取的图像边缘连续、平滑、完整,并有效地减少了图像边缘的冗余像素,进而提升了图像后续处理的效率;边缘跟踪方向抗干扰性强,具有较强的鲁棒性。为了减小检测的图像边缘与真实图像边缘之间的偏差、提高图像边缘检测的精度,参考边缘像素点的相邻区域灰度,以边缘像素点的梯度分布为依据对该像素点进行亚像素定位。经实验验证,经过亚像素优化的图像边缘检测策略可用于检测边缘模糊、对比度低的图像,检测的图像边缘完整、连续且平滑。该策略有效地消除了程序运算中引入的截断误差,提升了图像边缘检测精度,且适用于亮度5~100000lx的高动态成像场景中。