In this blog we will cover sample weed detection case study research paper. Our expert provide all research paper implementation assignment help, Case study assignment help, report writing assignment help which is related to data science or machine learning.
Weed Detection Sample Paper
Abstract – There are many restaurants opening in day to day. In this the Bengaluru has more than 12,000 restaurants serving dishes from all over the world. As per increasing demand in daily life there are many new restaurants opening. In this proje
ct report we will compare the performance of the customer data which is taken from Kaggle. Here our aim to analyse and select the best database (NoSQL and Relational) which is fit for big data.
Now a days online food delivering system is increasing as per customer demands so the data also increases as per customer order. So that it is necessary to analyse the data as per business purpose to get profit and loss.
Big data expert faces the issue to select database which is better for analyse big records which is make easy to execute and obtain the better result.
Today, many tasks in smart agriculture such as disease detection in plants, yield prediction, species identification, weed detection, and water and soil conservation are being solved using computer vision technology. Weed control is essential To increase the productivity of crops. Extensive literature has suggested precise variable spraying techniques to avoid the waste and herbicide residue problems caused by traditional full coverage spraying. In order to achieve accurate variable spreading, the key problem of how to realize accurate detection and identification of crops and weeds in real-time must be addressed.
How to detect field weeds using computer vision technology It mainly includes traditional image processing and deep learning.weed detection This is done using traditional image processing techniques that extract features such as the Combination of colors, textures, and shapes in images with traditional machine learning techniques such as random forest or support vector machine (SVM) algorithms for weeds ID is required. This method requires designing the elements by hand and High reliance on imaging technology, pre-processing technology, and quality feature extraction. With increased processing power and increased data volume Volumetric, deep learning algorithms can extract multidimensional and multidimensional spaces. Information on the semantic properties of weeds through Convolutional Neural Networks (CNNs) due to improved data representation on images while avoiding shortcomings. traditional extraction method.
Therefore, they are attracting more and more attention.
Weeds are one of the most important factors affecting agricultural production. Garbage Contamination of agricultural land from full-scale herbicide spraying is becoming increasingly evident. Continuously improve agricultural yields to accurately distinguish crops from weeds and achieve accurate weed-only spraying. important. However, the accurate application depends on the accurate identification and localization of weeds. and crops. In recent years, some scientists have been using various computer vision techniques for teeth. This review details two aspects of using existing image processing techniques. Deep learning techniques to solve the problem of weed detection. provides an overview Analysis of various weed detection methods in recent years, their pros and cons Familiarity with existing methods as well as several relevant plant leaves, weed datasets, and weeding mechanisms. Finally, we analyze the problems and difficulties of existing weed detection methods, The direction of future research development is predicted.
Traditional weed detection and deep learning techniques have their advantages. Traditional weed detection methods have small sample sizes,low GPU requirements, and Equipment at an affordable price. This article mainly deals with related weed detection methods. In recent years, traditional machine learning (ML) methods and Describes as deep learning and briefly discuss the pros and cons of the method. weed data set The challenges faced with the identification, detection, and classification of leaves are summarized. Weed detection of individual plots under various conditions was analyzed. in this article Definitive reference to other scientists to continue research on weed detection algorithms Based on advances in computer vision and intelligent weed control and related research and applications.
In the early days, many scientists used machine learning algorithms in combination with: We use image features to perform weed recognition tasks to achieve weed detection goals. These traditional machine learning methods require small sample sizes and short training times. Their GPU requirements are also low. They can be used in agriculture. We provide machines and equipment at an affordable cost, providing effective methods and approaches. Plant identification and weed detection based on imaging technology.
These intelligent technologies are based on the continuous development of machine vision. Technology. Machine vision technology uses a variety of image processing technologies to extract the shallow characteristics of the weed and then send it to a classifier for detection. Crops or weeds are initially identified by calculating their texture, shape, color, or spectral characteristics. Characteristics of the image. For example, Le et al. I realized the difference between corn and corn. Weed species were selected based on Local Binary Patterns (LBP) and SVM texture characteristics. Chen et al. We proposed a method for reverse finding multi-vertebral weeds in soybean fields. Based on shape and color characteristics. Zhu et al. Proposed classification method Five types of weeds on farmland based on their shape and texture. Zhang et al.
Comparatively analyzed the gray distribution of each component by color. RGB, HSV, and HIS space of common weeds in the field at the pea seedling stage. We proposed a method for segmenting and extracting weeds from complex backgrounds based on: About the nature of RB color difference. Some scientists have used the height or location of plants. Information to increase the accuracy of identification, but these methods In actual use, it may experience vibration or other uncontrolled movements. As well as, Some studies have focused on using single traits to identify low-level plants. Accuracy and low stability
Collection of Data
We are using a dataset uploaded to Kaggle. the data presented in this analysis are images of edible crops and weeds in the early stages of growing into seedlings. Recently, machine learning, deep learning, and image processing have shown great potential for advancing the digital capabilities needed for the future of agriculture. Robots and vision machines must be able to accurately identify and detect useful factory waste. Creating reliable weed/crop detectors in robotic weed control requires a focus on digital imaging-based sensing, mapping, management, and control technologies. The data provided in the dataset is for identifying food crops and weeds in the images.
Data Set link:
This dataset contains 1300 images of sesame crops and different types of weeds with each image labels.
Each image is a 512 X 512 color image. Labels for images are in YOLO format.
CNN Since the 1950s, when AI was in its initial stages, experts have tried to create a technology that can grasp visual input. This discipline became referred to As compubter Vision in the years that followed. When a team of academics from the University of Toronto built an Ai system that outperformed the top image recognition algorithms by a large factor in 2012, computer vision experienced a quantum leap. AlexNet was the name given to the Ai model. Convolutional Neural Networks, a form of neural network that approximates human vision, were at the core of AlexNet. CNNs became an increasingly crucial component of several computer vision applications, and thus a component of any web based vision course. So, let's have a look at how CNNs function. A convolutional neural network (CNN/ConvNet) is a type of deep neural network that is often used to evaluate visual images in deep learning. As we speak about neural networks, we tend to think of matrix multiplications, but that is not the issue for ConvNet. It employs a method known as Convolution. Convolution is a mathematical function between two factors that yields a third function that explains how the form of one is affected by the other. Bottom line, the ConvNet's function is to compress the pictures into a format that is simpler while retaining elements that are crucial for obtaining a successful forecast. Convolutional neural networks are made up of a number of levels of artificial neurons. Artificial neurons are computational models that determine the weighted sum of many inputs and output an activation value, which is an approximate replica of its biological counterparts. When you feed a picture into a ConvNet, each layer creates many activation functions, which are then transmitted on to the next layer. Typically, the first layer removes fundamental information such as lateral or longitudinal edges. This output is sent to the next layer, which recognises more complicated characteristics like corners or combinational edges. As we get further into the network, it can recognise more complex characteristics such as objects, faces, and so on. The classification layer generates a series of confidence ratings (numbers ranging from 0 to 1) depending on the activation map of the last convolution layer, indicating how probable the picture is to correspond to a "class." For example, if
you have a ConvNet that identifies cats, dogs, and horses, the last layer's result is the likelihood that the input image includes any of those creatures.
Pooling layer: The Pooling Layer, like the Convolutional Layer, is in charge of lowering the dynamic range of the Convolved Feature. By lowering the size, the computer power required to process the data is reduced. Average pooling and maximum pooling are the two forms of pooling.
Model Architecture the architecture of model is given below:
Optimizer: Adam is an adaptive learning rate optimization technique created particularly for deep neural network training. Adam, which was first reported in 2014, was exhibited at the renowned ICLR 2015 symposium for deep learning practitioners. To identify individual learning rates for every variable, the technique makes use of the capability of adaptive learning rate approaches. It also has the benefits of Adagrad, which performs well in situations with dense gradients but suffers in non-convex optimization of neural networks, and RMSprop, which attempts to alleviate some of Adagrad's difficulties and performs well in on-line scenarios. Adam's reputation has been skyrocketing.
Adam may be thought of as a hybrid of RMSprop and Stochastic Gradient Descent with momentum. It scales the learning rate utilizing squared gradients, similar to RMSprop, and it makes use of momentum by utilising the rolling average of the gradient rather than the gradient directly, similar to SGD with momentum.
Adam is an adaptive learning rate approach, which implies that it calculates specific learning rates for various variables. Its name is derived from adaptable moment estimation, and Adam employs estimates of the first and second moments of gradient to change the learning rate for each component of the neural network. Now, what exactly is the moment? The anticipated frequency of a random variable to the power of n is defined as its N-th moment. It is worth noting that the gradient of the cost function of a neural network can be regarded a random parameter because it is often assessed on a tiny randomized set of data. The first instant represents mean, while the second moment is uncentered variance (we don't remove the mean when calculating variance).
Binary_crossentropy: The Loss Function is a vital part of Neural Networks. Loss is anything but a Neural Net forecast mistake. And the procedure for calculating the loss is known as the Loss Function. To put it simply, the Loss is utilised to compute the gradients. Gradients are also utilised to modify the Neural Net's weights. This is the method through which a Neural Net is trained. For binary classification problems, BCE loss is employed. When utilising the BCE loss function, you only need one output node to divide the dataset into two groups. The output value must be processed through a sigmoid activation function, with a spectrum of (0 – 1). The binary cross entropy loss function is just suitable with the sigmoid activation function. Use a feature set to collect all of the Numeric characteristics that you really want the algorithm to estimate at the same time for the target block's Feature. Instead, you may evaluate the amount of each label with a separate Numeric feature that used a Numpy array without needing to establish a feature set.
Calculating Loss and Accuracy
Now we visualize the loss and accuracy of the models:
Color classification advances for longstanding weed identification, deep convolutional neural network, CNN-based techniques for isolating sugar beet plants and weeds, Gabor wavelet, Gabor wavelets and neural network, Kalman filter, decision trees and artificial neural networks, hyperspectral image analysis with wavelet analysis, support vector machine, Haar wavelet morph have all been suggested for weed classification. Skovsen et al. introduced completely CNN algorithms for weed and grass detection. Tang et al used the Kmeans function in conjunction with CNN to detect weeds. Adel et al. demonstrated the use of support vector machines and artificial neural networks to identify weed plants based on form data. These strategies have resulted in extraordinary performance and have shown outstanding progress in the agriculture industry. To increase the accuracy of classification of weed plants, however, efficient and sophisticated classification methods are necessary.
 S. H. Lee, C. S. Chan, S. J. Mayo, and P. Remagnino, "How deep learning extracts and learns leaf features for plant classification," Pattern Recognition, vol. 71, pp. 1-13, 2017
 S. G. Wu, F. S. Bao, E. Y. Xu, Y.-X. Wang, Y.-F. Chang, and Q.-L. Xiang, "A leaf recognition algorithm for plant classification using probabilistic neural network," in 2007 IEEE international symposium on signal processing and information technology, 2007: IEEE, pp. 11-16.
 Farooq, J. Hu, and X. Jia, "Analysis of Spectral Bands and Spatial Resolutions for Weed Classification Via Deep Convolutional Neural Network," IEEE Geoscience and Remote Sensing Letters, vol. 16, no. 2, pp. 183-187, 2019.
 J. Ishak, A. Hussain, and M. M. Mustafa, "Weed image classification using Gabor wavelet and gradient field distribution," Computers and Electronics in Agriculture, vol. 66, no. 1, pp. 53-61, 2009
 Ahmad, M. H. Siddiqi, I. Fatima, S. Lee, and Y.-K. Lee, "Weed classification based on Haar wavelet transform via k-nearest neighbor (k-NN) for real-time automatic sprayer control system," in Proceedings of the 5th International Conference on Ubiquitous Information Management and Communication, 2011: ACM, p. 17.
 Ahmed, A. Bari, E. Hossain, H. A. Al-Mamun, and P. Kwan, "Performance analysis of support vector machine and bayesian classifier for crop and weed classification from digital images," World Applied Sciences Journal, vol. 12, no. 4, pp. 432-440, 2011.
 W. Strothmann, A. Ruckelshausen, J. Hertzberg, C. Scholz, and F. Langsenkamp, "Plant classification with In-field-labeling for crop/weed discrimination using spectral features and 3D surface features from a multi-wavelength laser line profile system," Computers and Electronics in Agriculture, vol. 134, pp. 79-93, 2017
 S. Skovsen, M. Dyrmann, J. Eriksen, R. Gislum, H. Karstoft, and R. N. Jørgensen, "Predicting Dry Matter Composition of Grass Clover Leys Using Data Simulation and Camera-based Segmentation of Field Canopies into White Clover, Red Clover, Grass and Weeds," in International Conference on Precision Agriculture, 2018: International Society of Precision Agriculture.
 L. Tang, L. Tian, B. Steward, and J. Reid, "Texture-based weed classification using Gabor wavelets and neural network for real-time selective herbicide applications," Transactions of the American Society of Agricultural Engineers, vol. 99, p. 7035, 1999 · P. K. Goel, S. O. Prasher, R. M. Patel, J.-A. Landry, R. Bonnell, and A. A. Viau, "Classification of hyperspectral data by decision trees and artificial neural networks to identify weed stress and nitrogen status of corn," Computers and Electronics in Agriculture, vol. 39, no. 2, pp. 67-93, 2003.