The Multi-object Recognition Method Of Cluster Kiwifruits Based On Machine Vision

Information recognition is one of the key points of fruit-vegetable robots.The multi-object fruit recognition of kiwifruit in complex environments is a difficult point of information recognition.It is also an important factor for the efficient operation of kiwifruit picking robots and is also the key to the simultaneous picking of multiple arms of kiwifruit picking robots.It is the focus of this study to explore the identification methods of multi-object fruits of kiwifruit.By analyzing the growth status of clustered kiwifruit in a complex environment,the correct kiwifruit multi-object fruit image acquisition method can be used to reduce the interference of unrelated background.In this study,Kiwifruit pods were selected as identification points to study the multi-objective fruit recognition methods of kiwifruit in complex environments.Finally,the fruit pixel coordinates of kiwifruit were obtained.At the same time,the study of image processing segmentation software in the image preprocessing stage can provide convenience for the identification of multi-objective fruit of kiwifruit and also provide reference for the identification of other fruits.The main conclusions of this study are:(1)Research on multi-objective fruit recognition method of kiwifruit based on R-GThe growth characteristics of kiwifruit were analyzed.The kiwifruit had the characteristics of cluster growth and vertical distribution.Select the image of the multi-object fruit from the bottom to remove the interference from the ground weeds and trunks.The RGB color component of the fruit part in the kiwifruit image was detected and it was found that the R component(red)occupies the main part in the fruit,and the G component in the background leaves constitutes the main part(green).The RGB component analysis of the acquired image was performed using image Pro plus software.It was concluded that R component was greater than G component,G component was greater than B component in the RGB component of kiwifruit,and R component was in the background(part of fruit without kiwifruit).The difference between the G component and the B component does not vary greatly.A multi-objective fruit recognition method based on R-G for kiwifruit was proposed.Fruit calyx was used as identification feature points to achieve the separation of fruit target and background.Finally,the fruit calyx feature points were identified according to the fruit target.The correct recognition rate of the algorithm was verified by experiments.The results show that the accuracy of the algorithm is 88.86%.The average false-positive and false-negative rates are below 10%.Error analysis was performed on the identification results of each set of images tested.The reason for the large error is that the fruit is blocked by leaves,trunks,etc.,and some of the fruits are immature,resulting in the obstruction of the fruit during the image recognition process,and the fruit is wrongly divided into backgrounds.(2)Improved K-means clustering multi-object kiwifruit recognition methodImproved K-means clustering clustered multi-object kiwifruit recognition method is proposed,transferring the color space from the original color space to Lab color space,according to analyzing the characteristics of Lab color space,the color information is contained in the a channel and the B channel,it provides a convenient for the K-means cluster.The multi-objective kiwifruit is separated from the background,to complete the recognition of kiwifruit by using the method.The use of new templates and the Improved K-means algorithm can avoid the local optimal solution of the situation.Channel transform of the image: the image smoothing processing is completed by using the Gauss under the RGB channel,ensuring the correct recognition of kiwifruit and the kiwifruit calyx point accurately.The image information is grouped into three categories by using the K-means clustering in Lab color space: fruits,leaves and other background.The test and comparison of the R-G segmentation method,Otsu threshold segmentation method,the traditional K-means method and Improved K-means clustering algorithm identify the recognition rate.The test results show that the R-G segmentation algorithm,Otsu threshold segmentation algorithm,the traditional K-means clustering fruit recognition method and Improved K-means method are 80.04%,4.85%,89.73%,95.14%.(3)Research on obtaining method of kiwifruit calyx coordinatesThrough the method of introducing an empty matrix,using the transformation between the matrix and the empty matrix to extract the kiwifruit coordinates,and then obtaining the multi-object kiwifruit calyx by comparing the way of finding the centroid and using the largest bounding rectangle of the area to obtain the center of the rectangle.The results of analysis and verification by comparative experiments show that the recognition rate can reach 90.91%.At the same time,the reason that the identification error is analyzed is that there are gaps between the adjacent fruits that are similar to those of the fruit.It lays the foundation for the conversion between subsequent pixel coordinates and actual coordinates.The analysis of the acquisition error of the fruit-corner coordinate can provide reference for avoiding the error reduction during the picking process of the multi-arm kiwifruit picking robot.(4)Chromatic Aberration Segmentation Method Image Processing Software Segmentation SystemChromatic aberration segmentation is based on the RGB components of the image,establishing a color-difference segmentation model,separating or weakening the background noise,and providing convenience for image preprocessing.For the overall design of the system,the system module is divided into an image acquisition module and an image processing module.The image acquisition module includes an LED light light module and a Microsoft camera image acquisition module.The image processing module includes a grayscale image acquisition interface,histogram acquisition interface and color difference segmentation module.Image preprocessing includes acquisition of the image to be processed,grayscale processing,obtaining a grayscale histogram,median filtering of the image,obtaining an RGB component image of the image,establishing a color difference segmentation model,and obtaining a color difference segmentation effect map,the software The image pre-processing of kiwifruit multi-target fruit recognition plays a significant role.At the same time,the application of the color difference segmentation image processing software system in other fields is analyzed,and the overall effect is obvious.It also analyzes the problems in fruit segmentation.The reason for the presence of shadows in the target fruit segmentation in fruit segmentation is that the reflective part is white,and the R component,G component,and B component are close to each other,so the component after R-G component processing is performed.The value is close to 0,so a black shadow appears on the fruit.For the identification of fruit,in order to solve the problem of light reflection,an algorithm for removing the shadow can be designed or a suitable image acquisition angle can be found.This provides ideas for increasing the fruit recognition rate for fruit identification.The software system can also provide references for the identification of other fruits.