基于机器视觉的景区智能垃圾桶设计

Design of intelligent trash bin in scenic spot based on machine vision

随着我国经济快速发展,人们的生活水平不断提高,假期出游成为了首选,进而导致了越来越多的景区垃圾的产生,引发很多问题。为了解决垃圾外溢污染、垃圾难以正确分类及无接触扔垃圾的问题,设计了一种可进行垃圾分类与满溢报警等功能的智能垃圾桶,主要包括硬件电路设计与垃圾识别模型。智能垃圾桶应用光电开关与推杆电机的配合,当人靠近时,光电开关控制推杆电机推动垃圾桶桶盖打开,通过计算选定了推杆电机的型号为LA-T8-12-15-100/155-64;运用超声波传感器检测桶盖与桶内垃圾的距离,距离小于设定的距离时,通过GSM短信报警模块进行短信通知;运用OpenMV来进行垃圾分类,使用不同的设备采集每一类垃圾的数据集来训练模型,最后,通过Softmax激励函数将分类值转换成概率值;为了避免数据集样本过少导致的过拟合现象,采取了迁移学习和数据增强的方法,该方法训练出来的模型在OpenMVIDE软件中进行仿真试验,识别准确率达到90%以上,结果表明,可以有效地进行垃圾分类。经过原型机测试以上功能均可实现。

With the rapid development of China's economy and the continuous improvement of people's living standards,holiday travel has become the first choice,which has led to more and more garbage in scenic spots and caused many problems.To solve the problems of garbage overflow pollution,difficult to classify garbage correctly and throwing garbage without contact,an intelligent trash can with functions of garbage classification and overflow alarm is designed,which mainly includes hardware circuit design and garbage identification model.The intelligent trash can use the cooperation of photoelectric switch and push rod motor.When people approach,the photoelectric switch controls the push rod motor to push the trash can cover open.Through calculation,the model of push rod motor is LA-T8-12-15-100/155-64;The ultrasonic sensor is used to detect the distance between the barrel cover and the garbage in the barrel,and when the distance is lower than the set distance,the short message notification is carried out through the GSM short message alarm module;Using OpenMV to classify garbage,using different equipment to collect data sets of each kind of garbage to train the model,and finally transforming the classification value into probability value through Softmax excitation function;In order to avoid the over-fitting phenomenon caused by too few samples in the data set,the methods of transfer learning and data enhancement are adopted.The model trained by this method is simulated in OpenMVIDE software,and the recognition accuracy is over 90%.The results show that it can effectively classify garbage.The above functions can be realized after prototype test.