基于电磁式同步执行机构的鲜枣分级机设计与试验

为了提高鲜枣机器视觉分级机的分级精度和速度,该文研究设计基于电磁式同步执行机构的鲜枣机器视觉分级机,整机由辊轮输送链、拨板链条、电磁铁、同步传动链轮、STM32F407图像处理器和STM32F103控制动作执行的协处理器等组成。执行机构单元拨板与辊轮输送同步,采用灵敏度高的电磁铁吸附硅钢开关扣板执行动作。图像处理流程为图像颜色分量、Sobel算子滤波、大津法二值化、椭圆拟合提取颜色和大小特征值。样机试验结果:拨板链同步位差为0.413 mm,单元拨板击撞力为2.57 N,红枣表皮瞬间承受最大力为12.69 N;每帧图像处理时间为50 ms,执行机构动作时间为40 ms,辊轮输送速度为320 mm/s,红枣分级速度达20个/s,实现鲜枣5个大小等级和1个绿色等级的分级,分级准确率为93.7%,无损率为100%,最大产能达1 400 kg/h。设计的基于电磁式同步执行机构的鲜枣机器视觉分级机实现了鲜枣的高准确率、高速度、无损分级要求,为机器视觉红枣快速分级设备的发展提供重要的技术支撑和理论参考。

Design and experiment of synchronization actuator system based on electromagnetism for fresh jujube classifier

The production of the machine, which classifies small-sized fruits according to their images, depends on the speed of the image classification and the actuator. The existing classifier for fresh jujube has the issue that the efficiency of image processing doesn't match the timeliness of executive mechanism. And the speed of the actuator and the accuracy of positioning restrict the efficiency of classification. In order to improve the speed of classification of the image classifier for fresh jujube, an image classifier for fresh jujube is designed based on electromagnetic synchronous executor in this paper. The machine is composed of roller conveyor chain, dialing plate chain, electromagnets, synchronous drive sprocket, STM32F407 image processor and STM32F103 co-processor. The dialing plate chain of the actuator consists of 60 unit dialing plates. Chain transmission was adopted to ensure the synchronous precision with the roller conveyor chain. The dialing plate tension spring was buckled by the switch of the buckle. The material of the buckle was silicon steel, of which the hysteresis loop was narrow, which can enhance the sensitivity of electromagnetic triggering, and made the dialing action finished instantly. The location of fresh jujube corresponding to the unit plate, which would guarantee the alignment accuracy between the head of the dialing plates and fresh jujube. Pop-upstrength of the unit dialing plate was designed according to the peel hardness, the size and the weight of jujube. In image processing, G channel component, Sobel operator filtering, Otsu binaryzation, and eigenvalues extracted by ellipse fitting were adopted. A threshold value was used to calculate the percentage of the color of green, extract the size of feature value of the color and classify the image of green jujube. The algorithm was suitable for fast processing embedded system. The theory of algorithms and the test of image classification show that the synchronization error of chain transmission was 0.413 mm, the pop-up force of unit dial plate was 2.57 N, which was less than the damage strength (12.69 N) of the peel of red jujube, the time of each image processing was 50 ms, the time of the actuator for acting was 40 ms, the speed of roller conveying was 320 mm/s, the speed of classifying was 20 jujubes per second, the accuracy rate of classification was 93.7%, the intact rate was 100%, and the maximum capacity was 1400 kg/h. Results indicated that the sensitive trigger and the synchronous mechanism of synchronous electromagnet ensured the classification precision, and the touch force of dial plate was less than the damage stress of the red jujube skin. Also, the hardware design of image processing was an embedded system, which met the requirements of fast calculation through the algorithm of image processing and feature extraction. The running effect of the prototype achieves the purpose of fast, accurate and nondestructive classification.