Write a full project report: Fruit Detection System Using Neural Networking. (following this procedure -Abstract Introduction, Methodology,...

60.1K

Verified Solution

Question

Programming

Write a full project report: Fruit Detection System Using NeuralNetworking. (following this procedure -Abstract
Introduction, Methodology, Dataset, CNN-TensorFlow, ExperimentalResult, Conclusion, Future Work, Reference)
N.B. If you don't finish the answer here fully, after I cansubmit another part here like Dataset, CNN. Make sure you shouldanswer properly part by part.

Answer & Explanation Solved by verified expert

4.1 Ratings (567 Votes)

FRUIT DETECTION SYSTEM USING NEURAL NETWORKING1 AbstractThe aim of this project is to utilize the knowledge of DNN deepneural network or Neural Network to generate a model for FruitDetection The system is accurate fast and reliable fruitdetection system which is a vital element of an autonomousagricultural robotic platform it is a key element for fruit yieldestimation and automated harvesting Recent work in deep neuralnetworks has led to the development of a stateoftheart objectdetector termed Faster Regionbased CNN Faster RCNN We adaptthis model through transfer learning for the task of fruitdetection using imagery obtained from two modalities colour RGBand NearInfrared NIR Early and late fusion methods are exploredfor combining the multimodal RGB and NIR information This leadsto a novel multimodal Faster RCNN model which achievesstateoftheart results compared to prior work with the F1 scorewhich takes into account both precision and recall performancesimproving from 0807 to 0838 for the detection of sweet pepper Inaddition to improved accuracy this approach is also much quickerto deploy for new fruits as it requires bounding box annotationrather than pixellevel annotation annotating bounding boxes isapproximately an order of magnitude quicker to perform The modelis retrained to perform the detection of seven fruits with theentire process taking four hours to annotate and train the newmodel per fruit2 IntroductionAccording to sourcing skilled farm labour in the agricultureindustry especially horticulture is one of the mostcostdemanding factors in that industry This is due to the risingvalues of supplies such as power water irrigation agrochemicalsand so on This is driving farm enterprises and horticulturalindustry to be under pressure with small profit margins Underthese challenges food production still needs to meet the growingdemands of an evergrowing world population and this casts acritical problem to comeRobotic harvesting can provide a potential solution to thisproblem by reducing the costs of labour longer endurance and highrepeatability and increasing fruit quality For these reasonsthere has been growing interest in the use of agricultural robotsfor harvesting fruit and vegetables over the past three decadesThe development of such platforms includes numerous challengingtasks such as manipulation and picking However the developmentof an accurate fruit detection system is a crucial step towardfullyautomated harvesting robots as this is the frontendperception system before subsequent manipulation and graspingsystems if fruit is not detected or seen it cannot be pickedThis step is challenging due to various factors among which areillumination variation occlusions as well as the cases when thefruit exhibits a similar visual appearance to the background asshown in Figure 1 To overcome these a wellgeneralised model thatis invariant and robust to brightness and viewpoint changes andhighly discriminative feature representations are requiredFigure 1 Example images of the detection fortwo fruits a and b show acolour RGB and a NearInfrared NIR image of sweet pepperdetection denoted as red bounding boxes respectivelyc and d are the detection ofrock melonIn this work we present a rapid training about 2 h on a K40GPU and realtime fruit detection system based on DeepConvolutional Neural Networks DCNN that can generalise well tovarious tasks with pretrained parameters It can be also easilyadapted to different types of fruits with a minimum number oftraining images In addition we introduce approaches that combinemultiple modalities of information colour and nearinfraredimages with early and late fusion For the evaluation wedemonstrate both quantitative and qualitative results compared toprevious See Answer

Get Answers to Unlimited Questions

Join us to gain access to millions of questions and expert answers. Enjoy exclusive benefits tailored just for you!

Membership Benefits:

Unlimited Question Access with detailed Answers
Zin AI - 3 Million Words
10 Dall-E 3 Images
20 Plot Generations
Conversation with Dialogue Memory
No Ads, Ever!
Access to Our Best AI Platform: Flex AI - Your personal assistant for all your inquiries!

Become a Member