Method And Equipment For On-line Detection Of Grape Powder And Grain Size Based On Machine Vision

In recent years,with the rapid development of our economy and the general improvement of people's consumption levels,the grape industry has been greatly developed,and grape have become one of the most popular fruits in China.However,the low level of commercialization has always restricted the shift of the grape industry to scale and industrialization.Grapes is a spike-like structure,many fruit particles,soft fruits,and easy bruising,therefore,post-harvest processing of grapes is difficult.At present,the vast majority of areas in China still use manual grading methods and simple tools to classify them based on experience.Artificial grading is low in efficiency,high in cost and high in intensity,therefore,the market urgently needs on-line detection and grading technology and equipment of grape quality.In this paper,a type of grapes –Red Globe grapes is regarded as study object,machine vision technology is used to establish a online level model of grape powder adhesion rate and grape size,a grape online detection and grading hardware and software system is designed.The specific content is as follows:(1)The experimental platform of red tape image acquisition is established.An online grape image acquisition device is established on an experimental circular transport line,an image acquisition device with a size of 600×450×450mm3 is designed.The grooving at the top of the device ensures the constant transport of the clamping mechanism,the photoelectric sensor in the dark box senses the grape and triggers the camera,thus realizing the continuous collection of red-lift images.Analyze and contrast the advantages and disadvantages of many different light sources,and determine the lighting type of the shooting device;In order to extract the morphological outline and color features of grape at the same time,the camera and lens that can simultaneously acquire the color image and the near infrared image is selected.(2)The discrimination method of attach rate grade of grape fruit powder is studied.Because disturbing green stemming information will intefere grape image,if color information extracted ways is used directly,the segmentation effect will not be good.In this image preprocessing stage,the normalized ultra green method is used to effectively remove the interference.The color feature values of grape powder and non-grape powder regions are extracted and classified by Naive Bayes,BP neural network and SVM model.Through comparison,it was found that the SVM model has a short detection time and high prediction accuracy.It is feasible to apply online identification of grape powder.By comparing the results of artificial classification,the classification accuracy of the grape fruit powder attach rate was 93%.(3)The method of detecting and grading grape size has been studied.A morphological method based on open-close reconstruction and local maximum is adopted to acquire the spot position on the surface of each fruit grain in the near infrared image,acquire the coordinates of the center position of the fruit grain,and complete the coarse positioning of the fruit grain.Gradient segmentation algorithm was used to obtain the true outline pixels of each grain.However,some overlapping contours seriously interfere with the target grain,so a random least square ellipse fitting algorithm was designed to extract the target grain size,then,complete the precise positioning of grapes.After verification with the grading results of manual measurement,the accuracy rate of the grape size detection and grading method designed in this study is 91%,realizes the automatic grading of the whole bunch of grapes.(4)For on-line detection and grading of grapes powder and size,an on-line control system based on machine vision and detection software has been designed.The on-line control system mainly includes the motor control system and the image acquisition control system,which realize the real-time control of the speed of the circular transport line and the online acquisition and processing of the red picture.The on-line detection software is designed using the MFC control in the VS software.The main functions of the software include the setting of the protocol in the upper computer,the display of the collected grape picture,the monitoring of the state of the serial communication and getting the result of grape powder and the size model.(5)A online detection and grading equipment has been designed.The pipeline equipment mainly includes a conveying device,an image acquisition device,a staged kickout device and a sorting device.The conveying device adopts the design method of the annular suspension conveyor chain,which ensures the stability of grape transportation and simulates the actual production application scenario.A clamping mechanism with a ratchettooth self-locking function has been designed to facilitate red-loading and loading and unloading processes and reduce manual labor.The step-out kicking device which has a fixed gap shape realizes the automatic and continuous sorting of grape;the designed sorting device composed of a plurality of belt rollers realizes the grapoe rating of the corresponding level.