Guidance of groups of guppies (Poecilia reticulata) to allow sorting by computer vision

Prior to marketing, ornamental tropical fish are currently sorted, graded and counted manually; a process that is labor intensive, stressful to fish and inaccurate. Manipulation of fish movements at our will is a key element in the development of a device for automatically sorting, grading and counting fish by a computer vision system, and the present study is one step towards the development of such a fish-friendly sorting device. The positive phototactic and rheotactic innate responses of guppies (Poecilia reticulata) were exploited to induce them to swim through transparent pipes or narrow channels, to render them ready for inspection by a computer vision system. Water flow direction and velocity, water level, and illumination were effective in inducing fish to move from one container into the other via a narrow transparent pipe. It was attempted to separate guppies moving in groups into individually spaced fish, by introducing an obstacle into a narrow channel or narrowing the transparent pipe.     

Color vision, spectral sensitivity, accommodation, and visual acuity in juvenile masu salmon Oncorhynchus masou masou

ABSTRACT:   Color vision, spectral sensitivity, accommodation, and visual acuity were examined in juvenile masu salmon Oncorhynchus masou masou to obtain fundamental information about the visual system. Two types of S-potentials were recorded from 415 horizontal cells in isolated retinas from 34 cultured freshwater masu salmon (114–219 mm standard length, SL). Although horizontal cells recording S-potentials were not identified, the horizontal cells were confirmed because their responses were maintained while the stimulus remained. The variety of chromaticity (C)-type S-potentials indicated well-developed color vision. The analysis of the luminosity (L)-type S-potentials indicated that the peak spectral sensitivity was at 522 nm. S-potentials were also recorded in response to ultraviolet light. The direction and extent of lens movement induced by electrical stimulation was measured in 12 cultured masu salmon (99.0–142.5 mm SL). The results indicated that the visual axis was upward and forward, and that the range of accommodation was from 0.79 × SL in front of the eye to infinity. In histological analysis of the retinas of five wild smolts (100–118 mm SL), the maximum cone densities (276–345 cones/0.01 mm²) were detected in the ventral to temporal regions. The visual acuities assessed by histological methods were 0.069–0.075.     

Behavioral responses of tilapia (Oreochromis niloticus) to acute fluctuations in dissolved oxygen levels as monitored by computer vision

Behavioral and ventilatory parameters have the possibility of predicting the stress state of fish in vivo and in situ. This paper presents a new image-processing algorithm for quantifying the average swimming speed of a fish school in an aquarium. This method is based on the alteration in projected area caused by the movement of individual fish during frame sequences captured at given time intervals. The image enhancement method increases the contrast between fish and background, and is thus suitable for use in turbid aquaculture water. Behavioral parameters (swimming activity and distribution parameters) and changes in ventilation frequency (VF) of tilapia (Oreochromis niloticus) responded to acute fluctuations in dissolved oxygen (DO) which were monitored continuously in the course of normoxia, falling DO level, maintenance of hypoxia (three levels of 1.5, 0.8 and 0.3 mg l⁻¹) and subsequent recovery to normoxia. These parameters responded sensitively to acute variations in DO level; they displayed significant changes (P < 0.05) during severe hypoxia (0.8 and 0.3 mg l⁻¹ level) compared with normoxic condition, but there was no significant difference under conditions of mild hypoxia (1.5 mg l⁻¹ level). There was no significant difference in VF between two levels of severe hypoxia 0.8 and 0.3 mg l⁻¹ level during the low DO condition. The activity and distribution parameters displayed distinguishable differences between the 0.8 and 0.3 mg l⁻¹ levels. The behavioral parameters are thus capable of distinguishing between different degrees of severe hypoxia, though there were relatively large fluctuations.     

种猪体重测量新方法初探

本文简述了体重测量在种猪饲养生产管理中的必要性,通过分析种猪各个体尺与体重之间的关系,探讨了一种无接触式的新测量方法,即用计算机视觉技术无接触地来估算体尺和预测体重,这种方法可以解决了传统测量方法的弊端,减少人力物力,避免由于猪的应激反应而给生产带来的损失,在种猪的科学饲养管理中具有重要的现实意义。     

基于改进YOLOV5s网络的奶牛多尺度行为识别方法

奶牛站立、喝水、行走、躺卧等日常行为与其生理健康密切相关,高效准确识别奶牛行为对及时掌握奶牛健康状况,提高养殖场经济效益具有重要意义。针对群体养殖环境下奶牛行为数据中,场景复杂、目标尺度变化大、奶牛行为多样等对行为识别造成的干扰,该研究提出一种改进YOLOV5s奶牛多尺度行为识别方法。该方法在骨干网络顶层引入基于通道的Transformer注意力机制使模型关注奶牛目标区域,同时对奶牛多尺度行为目标增加路径聚合结构的支路与检测器获取底层细节特征,并引入SE（Squeeze-and-Excitation Networks）注意力机制优化检测器,构建SEPH（SE Prediction Head）识别重要特征,提高奶牛多尺度行为识别能力。试验验证改进后的奶牛行为识别模型在无权重激增的同时,多尺度目标识别结果的平均精度均值较YOLOV5s提高1.2个百分点,尤其是对奶牛行走识别结果的平均精度4.9个百分点,研究结果为群体养殖环境下,全天实时监测奶牛行为提供参考。     

基于中值点Hough变换玉米行检测的导航线提取方法

为解决机器视觉对早期玉米苗带在多环境变量下导航线提取耗时长、准确率低的问题,该研究提出了一种基于中值点Hough变换作物行检测的导航线提取算法。首先,改进了传统的2G-R-B算法,再结合中值滤波、最大类间方差法和形态学操作实现土壤背景与玉米苗带的分割。其次,通过均值法提取玉米苗带特征点,然后采用中值点Hough变换拟合垄间两侧玉米苗列线,最后将检测出的双侧玉米苗列线为导航基准线,利用夹角正切公式提取导航线。试验结果表明：改进的灰度化算法能够正确分割玉米苗带与土壤,处理一幅640×480像素彩色图像平均耗时小于160 ms,基于中值点Hough变换检测玉米苗列再提取导航线的最大误差为0.53°,相比于传统Hough变换时间上平均快62.9 ms,比最小二乘法平均精确度提高了7.12°,在农田早期玉米苗带多环境变量影响因素下导航线提取准确率均达92%以上,具有较强的可靠性和准确性。     

无人机点射式水稻播种装置控制系统设计与试验

针对当前无人机水稻撒播难以成行成穴、落种易受旋翼风场干扰和播种均匀性不佳等问题,该研究结合点射式水稻播种装置和飞行控制器设计了一套播种控制系统,开发了配套的地面站功能,并制作了样机。控制系统基于PID算法实现排种器步进电机的转速闭环控制,通过标定模型对振动电机激振力和摩擦轮电机转速进行控制,并根据状态机设计播种控制程序。以3倍丸粒化稻种为对象,从播种量准确性、播种成行性和播种均匀性3个方面对样机的播种性能进行验证并优选合适的播种参数。试验结果表明：无人机模拟飞行的播种量准确性测试中,样机以1.0～2.5 m/s的作业速度进行播种时,播种量的平均相对误差小于4%,控制系统具有较好的动态调节能力。实地飞播测试中,样机以1.0和1.5 m的高度播种时,种子分布在12 cm种行宽度内的平均概率超过80%,成行性较好。考虑安全因素,优选1.5 m为样机的适宜作业高度。在作业高度为1.5 m,3倍丸粒化稻种的播种量为90～150 kg/hm2（对应裸种的播种量22.5～37.5 kg/hm2）,作业速度为0.5～2.0 m/s时,播种均匀性变异系数为20.51%～35.52%。进一步分析发现,适当提升作业速度可提高播种均匀性。田间试验结果表明,播种量的相对误差分别为2.47%和4.12%,播种均匀性变异系数分别为22.17%和21.82%,种子破损率分别为0.34%和0.18%,满足相关标准的水稻飞播精度控制要求。研究结果可为无人机水稻直播技术提供参考。     

苹果外部缺陷全表面在线检测分选装置研发

中国是水果消费大国,但在水果产后检测装备方面相对滞后。针对目前在线装置无法采集苹果全表面图像信息且无法精确计算缺陷面积的问题,该研究以表面缺陷面积的快速检测为主要目标,提出苹果全表面图像合成算法,设计了一套苹果外部品质在线检测及分级装置。该研究以苹果为例,基于球模型提出苹果全表面图像合成算法、缺陷面积校正算法精确计算苹果的表面缺陷面积。通过试验验证,对苹果表面图像进行分割合成后,整体的图像的漏检率为0。提出缺陷面积校正算法,可以计算图像中位于任意位置的苹果缺陷真实面积,选取了120个样本进行验证,其中擦伤样本、碰伤样本、痘斑病样本、表面腐败样本各30个。擦伤样本缺陷面积预测值与真实值的决定系数R2为0.9787,标准误差RMSE（Root Mean Squared Error）为3.577 4 mm2,偏角试验中R2为0.975 8,RMSE为3.466 3 mm2。碰伤样本缺陷面积预测值与真实值的R2为0.973 0,RMSE为3.981 9 mm2,偏角试验中R2为0.974 2,RMSE为4.062 4 mm2。痘斑病样本缺陷面积预测值与真实值的R2为0.970 8,RMSE为3.836 6 mm2,偏角试验中R2为0.977 9,RMSE为3.895 3 mm2;表面腐败样本缺陷面积预测值与真实值的R2为0.9812,RMSE为3.178 1 mm2,偏角试验中R2为0.974 8,RMSE为6.304 4 mm2。在整个试验过程中,R2总体上高于0.97,RMSE小于6.304 4 mm2。装置检测苹果的速度为2个/s,评级准确率为95%,检测与苹果评级精度较高,工作较为稳定,实现了苹果外部缺陷的检测与分级评价,为苹果的外部品质检测提供了技术支撑。     

基于模糊聚类分析的草莓图像分割方法研究

为改善草莓采摘机器视觉系统中果实图像的分割效果,对普通均值聚类的分割方法理论进行分析,针对草莓果实图像的特点将模糊-均值聚类算法引入分割算法,大大改善草莓果实图像的分割效果。     

机器视觉在植物叶片叶面积测算中的应用

利用机器视觉系统代替人工对叶片叶面积进行测算。运用Halcon图形开发工具,以VB 2008为基础开发平台,采用通过阈值分割与区域特征提取的方法来计算叶面积,并将结果显示到屏幕并保存到文件中。将具体图像采集设备与机器视觉系统连接,完成叶片图像采集、图像处理和结果显示的一体自动化。与手工测算的费时费力和使用专业设备的高昂费用相比,此系统能在保证测算结果精确度的基础上,对多种植株叶片进行快速准确、简单易行、经济实用的测算。