Smart Fruit Basket: Towards Multi-View Fruit Recognition

Smart Fruit Basket: Towards Multi-View Fruit Recognition

Pulkit Narwal, Ipsita Pattnaik
Copyright: © 2022 |Pages: 14
DOI: 10.4018/IJCICG.311427
OnDemand:
(Individual Articles)
Available
$37.50
No Current Special Offers
TOTAL SAVINGS: $37.50

Abstract

This paper discusses smart retailing solutions, self-checkout stores in particular. Since RFID tag-based product identification accounts for various limitations, the authors propose a smart basket to facilitate self-checkout mechanism for fruits and vegetables, based on multi-view image recognition and weight sensor. The system works on a multi-view model and recognizes and counts the fruit/vegetables from four camera views to handle the occlusions. The user places fruits inside the basket. Multiple cameras installed provide different views inside the basket and captures this fruit placing activity. Different views are then processed for image recognition using CNN (convolutional neural network). The authors also present a multi-view fruit recognition (MVFR) dataset to evaluate the system performance. The base of smart basket includes a weight sensor to account for weight information, the weight, and count information of fruit assist in bill generation at self-checkout station.
Article Preview
Top

Introduction

In recent years, with the advancement of Artificial Intelligence, the retail industry is changing like never before. Inclusion of computer vision, Radio Frequency Identification (RFID), machine learning, and sensors has reduced the requirement of manual work at its minimum. Among such trends, Vision based retailing has seen some great changes. Be it unmanned stores, smart unmanned vending machines, self-checkout stations or unmanned shelf monitoring. Self-checkout stations in retail stores are incorporating new vision based techniques to reduce manned task, and assist users to self-checkout.

This paper works on fruit recognition based on Convolutional Neural Network (CNN). The paper focusses on providing self-checkout mechanism for fruits and vegetables through object recognition. Inspired by various self-checkout facilities provided, the need of manual work at billing or checkout counters is substituted by smart solutions. Though, there has been some significant research contributions for user centric self-checkout techniques including one dimensional barcodes, two dimensional Quick Response (QR) codes and RFID approaches, they are still hindered by the limitations of requirement of RFID tags, QR codes and barcodes on the product. But employing these tags or codes on every product may not be a viable option especially with fruits and vegetables. Thus, existing approaches may not work, given there is no code or tags attached to the product. The authors hereby propose a machine learning based approach to recognize the fruits built on top of computer vision framework. The system works on a multi view model and recognizes and count the fruit/vegetables from multiple views provided by different camera views. The occlusions are well handled by a multi view fruit-vegetable recognition strategy. The purpose of self-checkout does not finish on recognition task, there is also a need of product counting and weighing mechanism. Thus, a smart basket is proposed, installed at checkout counters to easily recognize the fruit and also determine the count of fruits along with weight. The detections provided by smart basket serve as a basis for bill generation on screen installed next to smart basket.

Fruit harvesting sector is revolutionized with new technologies, ranging from harvesters and other machines to robotic harvesting. As these methods require pay-per-quantity and pay-per-hectare payment models. Fruit farmers generally hire external harvesting agencies for harvesting. These external harvesting agencies charge per hectare of the harvest. Also, potential buyers of fruits makes an agreement with fruit farmers and harvest on their part and later buy the fruit harvest. This agreement generally resorts to pay-per harvest or pay-per quantity payment models. In pay-per-harvest, a price is fixed per hectare of harvest land and the fruit farmer is entitled to the payment based on total hectare of harvest land times the price fixed. Whereas, in pay-per-quantity model, a price is fixed per quantity of fruit and the farmer is entitled to the payment based on total quantity of fruit times the price fixed. It becomes important to predict the harvest beforehand and use the services accordingly. Also, planning before fruit ripening with harvest estimation proves beneficial. This work can also be applied to predict the fruit harvest from UAV (Unmanned Aerial Vehicle) images and videos.

Complete Article List

Search this Journal:
Reset
Volume 14: 1 Issue (2024): Forthcoming, Available for Pre-Order
Volume 13: 2 Issues (2022): 1 Released, 1 Forthcoming
Volume 12: 2 Issues (2021)
Volume 11: 2 Issues (2020)
Volume 10: 2 Issues (2019)
Volume 9: 2 Issues (2018)
Volume 8: 2 Issues (2017)
Volume 7: 2 Issues (2016)
Volume 6: 2 Issues (2015)
Volume 5: 2 Issues (2014)
Volume 4: 2 Issues (2013)
Volume 3: 2 Issues (2012)
Volume 2: 2 Issues (2011)
Volume 1: 2 Issues (2010)
View Complete Journal Contents Listing