论文部分内容阅读
[目的]新生仔猪目标检测是母猪分娩监测的关键环节。[方法]通过自制图像采集器采集母猪分娩视频图像,机器视觉系统获取分娩图像信息,选取Canny算子对图像进行边缘检测,采用Otsu算法对图像进行二值变换,应用滑动平均算法和形态学开运算对二值图像滤波消噪,提取图像最大连通域,利用团序列检测算法对母猪目标进行分割,对分割后区域进行仔猪目标识别。[结果]试验结果表明,团序列检测算法能够准确分割出母猪目标,检测仔猪目标的正确率达到95.5%。[结论]提出一种能够有效识别新生仔猪目标的方法,为仔猪的出生预警提供了技术支撑。
[Objective] Newborn piglet target detection is the key link of sow birth monitoring. [Method] The video images of farrowing sow were collected by self - made image acquisition system. The machine vision system acquired the image information of delivery. The Canny operator was selected to detect the edge of the image. Otsu algorithm was used to transform the image binary. The sliding average algorithm and morphological Open operation filters and denoises the binary image, extracts the maximal connected region of the image, uses the group sequence detection algorithm to segment the sow target, and performs the target recognition to the segmented region. [Result] The test results showed that the algorithm of group sequence detection can accurately segregate the target of sow, and the correct rate of target detection of piglets reached 95.5%. [Conclusion] A new method that can effectively identify the target of newborn piglets is provided, which provides technical support for early warning of piglets.