基于嵌入式的双目视觉采棉测距系统

Embedded binocular vision cotton picking distance system

针对当前棉花采摘劳动量大、采摘机器效率普遍低下、成本较高的问题,研发并设计了一种基于嵌入式的双目视觉采棉测距系统,以提取棉花采摘过程中棉花的纵向距离。系统使用双目立体摄像头采集图像数据,以ARM为核心,将电源、通信、内存扩展和显示调试等功能模块分布在外围,实时显示图像。在嵌入式系统中,对采集到的棉花图像进行二值化、中值滤波等预处理,运用基于区域生长的广度优先标记算法,检测并标记出目标棉花所在的连通区域,实现了特征点的匹配和中心采摘点的提取。实验测得目标棉花到摄像头在300~800 mm范围内的距离,测量误差在10%以下。结果表明,该系统能够实现对棉花图像的采集、显示、识别处理和纵深测距,测量结果合理,工作稳定(准确性和可靠性),具有良好的人机交互功能和跨平台移植性。最大的优点是脱离计算机,操作方便、体积小且成本低,对比高精度设备,精度达到了要求,具有一定的实际意义和应用价值。

In view of the current large amount of cotton picking labor, the efficiency of picking machines is generally low, and the cost is high, the paper developed and designed an embedded binocular vision cotton picking distance system to extract the longitudinal distance of cotton during cotton picking. The system uses a binocular stereo camera to capture image data. With ARM as the core, functional modules such as power supply, communication, memory expansion and display debugging are distributed around the periphery to display images in real time. In the embedded system, the collected cotton images are preprocessed by binarization and median filtering, and the breadth-first marking algorithm based on region growth is used to detect and mark the connected region where the target cotton is located. Thereby, the matching of the feature points and the extraction of the center picking points are realized. The distance between the target cotton and the camera in the range of 300~800 mm is measured by the experiment, and the measurement error is below 10%. The results show that the system can realize the collection, display, recognition processing and depth ranging of cotton images, the measurement results are reasonable, the work is stable(accuracy and reliability), and it has good human-computer interaction function and cross-platform portability. The biggest advantage is that it is separated from the computer, easy to operate, small in size and low in cost. Compared with high-precision equipment, the precision meets the requirements, and it has certain practical significance and application value.