基于BP网络的活塞裙部纵向型线的拟合

在分析有关活塞裙部纵向型线拟合方法的基础上 ,提出了基于 BP网络拟合的方法 ;应用实例拟合结果表明 :基于 BP网络的拟合方法适用性广 ,且拟合精度较高。     

基于轮廓分析的双串叠贴葡萄目标识别方法

为准确定位叠贴情况下的葡萄目标,提出了一种基于轮廓分析的双串叠贴葡萄目标识别方法。首先提取最能突显夏黑葡萄的HSV颜色空间中的H分量,通过改进K-means聚类方法对葡萄图像进行分割,运用形态学去噪等处理获取葡萄图像区域,再提取该区域边缘轮廓和左右轮廓的类圆中心。然后以该中心点为原点建立基于轮廓分析的叠贴葡萄串分界线几何求解与计算模型,分别在逆时针方向45°~135°和225°~315°区域内沿葡萄轮廓搜索距离原点最近的点,进而确立两叠贴葡萄轮廓拐点及其分界线,最终实现对叠贴葡萄目标的分别提取。对从果园采集的27幅双串叠贴葡萄图像进行试验,结果显示:24幅图像中的叠贴葡萄串被正确识别和提取,成功率达88.89%,目标像素区域的识别精准度为87.63%~96.12%,算法处理时间在0.59~0.68 s之间。将算法移植到自主研制的机器人上进行视觉定位试验,结果表明所提方法可很好地用于两叠贴葡萄目标的识别与定位。     

基于图像处理的生猪耳部颜色异常检测技术

为研究规模化生猪养殖场中蓝耳病疫情预警监测方法,提出一种非接触式耳部颜色自动检测方法。该方法将生猪热红外图像和可见光图像相结合寻找最优尺度因子,确定可见光图像中生猪耳根部特征区域;采用主动形状模型方法,选取34个生猪耳部轮廓特征点,并将搜索范围限定在生猪头部区域,用以提取生猪耳部轮廓;将提取的耳部轮廓进行颜色对比,判断该生猪是否患有蓝耳病疫情。结果表明,由于限定生猪头部区域搜索范围,能快速准确地提取生猪耳部轮廓。对生猪耳部颜色检测准确率达到77%以上。     

基于主动轮廓模型的自动导引车视觉导航

为解决工厂或农业非线性光照环境下视觉导航算法鲁棒性和实时性差的问题,提出基于主动轮廓模型的导航标识线检测与跟踪算法。首先用多项式曲线模型描述单向标识线,导航问题等价为计算多项式曲线参数;然后依据标识线颜色和边缘等特征,给出关于多项式曲线的主动轮廓模型的内、外部能量函数;最后将能量函数简化为非线性最小二乘问题,应用高斯牛顿法和Armijo-Goldstein不精确一维搜索方法求解曲线最优参数。采用自制视频和自主小车测试算法,结果表明:该算法对非线性光照条件下直线和弯曲标识线的导航正确率为98.96%,运算时间为40.18 ms。试验验证了该算法的鲁棒性和实时性。     

Visual Navigation for Automatic Guided Vehicles Based on Active Contour Model

Lane detection and tracking algorithm based on active contour model was proposed to solve the poor robustness and real time problem for vision navigation under factory or agricultural nonlinear illumination conditions. First of all, it was illustrated that navigation problem was equivalent to calculation of polynomial curve parameters, which could describe the navigation lanes. Secondly, the external energy function of active contour model was investigated, including three energy terms. The first energy term was about the Euclidean distance between lane colors and colors on one side of polynomial curve, by minimizing the first energy term could attract polynomial curve to navigation lanes. The second energy term was about the edge features, which could attract polynomial curve to lane edges. The third energy term was about the position difference of polynomial curve between adjacent frames, which could limit curve to change abruptly. Finally, the energy function was simplified to a nonlinear least squares problem, and the Gauss-Newton method as well as the Armijo-Goldstein inexact line search method were used to solve this problem. Home video and independent car were tested, the result showed that the algorithm achieved a navigation accuracy of 98.96% for both the straight lane and bending lane under nonlinear illumination, with average processing time of 40.18ms, and the independent car could walk along the navigation lane successfully. Experiment result showed that the algorithm was robust and real-time.