【作者】 张海亮；

【导师】 何东健；

【作者基本信息】 西北农林科技大学 ， 机械设计及理论， 2003， 硕士

【摘要】 肉牛业在中国的出现是农村改革开放成功的标志。随着人民生活水平的提高，肉牛业在农业生产和国民经济中的地位显得越来越重要，自动化和产业化是肉牛业的必然趋势。研究针对肉牛体型参数人工测量效率低，而且测量精度不高，测量结果易受测量者技术水平、疲劳程度和测量仪器本身精度的制约等问题，应用计算机视觉技术进行肉牛主要体型参数的测量，并基于Web对图像信息进行管理。主要研究内容如下： (1)根据研究性质和要求，分析了影响系统性能的因素，综合分析性能价格比的基础上，组成了计算机视觉系统。选择合适的摄像角度，在自然环境下获取肉牛图像。 (2)运用参考物标记法对体型参数进行测量，使得图像获取过程不受摄像焦距的影响，使图像大小和清晰度得以改善，给测量带来了很大方便。 (3)利用判别分析法确定阈值，色度能有效的识别肉牛和背景，识别正确率在98％以上，且对土壤、光照强度变化、建筑物阴影的敏感性小。 (4)以区域面积为依据，去除测量对象以外的无用信息、噪声信息。在色度二值化后，无需通过其他去噪方法，也可以准确的去除无用信息，从而大大提高了目标识别的效率。 (5)针对获取的二维图像在测量时，有些参数无法直接测量的情况，本文提出通过近椭圆计算和人工神经网络数据拟合法，通过胸宽和胸深实现了对胸围的计算。 (6)在参数测量中，通过分析肉牛形体特征，通过局部最大曲率以及考察点相对坐标夹角变化能够较为准确的检测测量点。 (7)基于Visual C++6.0开发了计算机视觉肉牛体型参数测量软件，能够实现重要体型参数的测量和主要体尺指数的计算，其基本的图像处理功能有较好的通用性。 (8)以B/S为模型设计了的图像信息管理系统，能够实现图像信息的常规管理(图像信息的添加、删除、更新和查询)，局域网测试结果表明，系统有良好的运行稳定性。更多还原

【Abstract】 The appearance of cattle-raising industry in china is the symbol of the success of revolution and opening. With the improvement of people’s living situation, the status of cattle-raising industry is becoming more and more important hi agricultural manufacture and national economy. Industrialization and automation are the inevitable trend. Aim at the problems of manual measuring of cattle’s shape parameters: low efficiency, low precision, the results easily effected by the technique level and fatigue level of worker, and the precision of measure tools, the research used computer vision technology to measure the main shape parameters of cattle. And the research setup a information management system to manage image information. The main contents and results of the dissertation as follows:(1) According to the property and demand of the study, the factors affecting system performance were analyzed. On the basis of considering the ratio of performance and price, hardwares were chosen to setup the computer vision system. Under the natural environment, cattle images were acquired.(2) Using reference object demarcation method to measure the shape parameters, this made the process of image acquiring avoid the effection of camera’s focus. And it can eveluate the image’s size and definition, at the same time it mad the measurment easy.(3) Hue can identify cattle from backgroud through Ostu’s thresholding, the correctness is over 98%. And it’s insensitive to earth, the illumiation intensity and shadow.(4) Basing of the area, can delete the area which is useless to measurment and noise. After binary based on hue, just use this method can delete useless information, then improved the efficiency of identifying.(5) Aimed at some parameters cann’t measure directly from the 2-dimension image, the paper indicate that can through near-ellipse calculation and using BP neural network to calculation the chest circumference of cattle.(6) In the process of measurment, when measure the shape parameter, can calculate the max local-curvature, and analysis the angle from the edge-point to coordinate system to find the beginning point and stopping point of measure.(7) Software system for cattle’s shape parameter measurment by computer vision were developed by using VC++6.0. It can measure the main shape parameters and main index from cattle image. The function of basic image processing can used in otherobjects and fields.(8) Designed and developed the information management system for cattle’s image based on the B/S modle. It can complete the basic function of information management(insert image information, delete image information, update and query image information). The system is running on local intranet to test it’s performance, and it works well.更多还原