Description

ABSTRACT

Fruit counting is an important task for growers to estimate yield and manage orchards. An accurate automated fruit detection and counting algorithm gives agricultural enterprises the ability to optimize and streamline their harvest process. Through a better understanding of the variability of yield across their farmlands, growers can make more informed and cost-effective decisions for labor allotment, storage, packaging, and transportation. Estimation of fruit count from images is a challenging task for a number of reasons including appearance variability due to illumination, and occlusion due to surrounding foliage and fruits.Fruit counting algorithms relied onOpen computer vision methods involving hand-crafted features that exploited the shape, color, texture or spatial orientation of various fruit. A counting algorithm based on a second convolutional network then estimates the number of fruit in each region. Finally, color maps that fruit count estimate to a final fruit count.This method generalizes

INTRODUCTION

Fruit counting is an important task for growers to estimate yield and manage orchards. An accurate automatedfruit detection and counting algorithm gives agricultural enterprises the ability to optimize and streamline their harvestprocess. Through a better understanding of the variability ofyield across their farmlands, growers can make more informedand cost-effective decisions for labor allotment, storage, packaging, and transportation. Smart sensor suites as well asautonomous robots such as unmanned aerial vehicles (UAVs)will benefit from data-driven fruit counting algorithms thatenable growers to estimate yield at scale across both data sets and is able to perform well even on highly occluded fruits that are challenging for human labelers to annotate.

EXISTING SYSTEM

• Fruit counting algorithms relied on traditional MATLAB based process
• Manually Counting
• Image segmentation Can be done

DISADVANTAGES

• It will take high processing Time
• Recognize only Fixed shape, color, texture

PROPOSED SYSTEM

• Open CV based Object Recognition and Counting
• Deep learning-based Approach

BLOCK DIAGRAM

                                 

BLOCK DIAGRAM EXPLANATION

The system consists of Cortex A-53 Raspberry Pi device and camera. Here camera used to capture the image frames; then, it will compare to the related fruits. If its recognized means count is increased one. Image reorganized process based on Open computer vision Algorithms.    

HARDWARE REQUIREMENTS

• Raspberry pi
• SD card
• Camera

SOFTWARE REQUIREMENTS

• Raspbian Jessie OS
• OpenCV
• Language – Python

CONCLUSION

We have presented a novel data-driven end-to-end fruitcounting pipeline based on deep learning that generalizesacross various unstructured environments. In order to demonstrate this generalization, we chose data sets that are challenging in unique ways: the orange data set features high level ofocclusions, depth variation, and uncontrolled illumination, andthe apple data set features high color similarity between fruitand foliage.