Plant seedling classification is crucial for biodiversity conservation. Based on analysis of CNN, we propose a system to classify plant seedlings with minimum classification error. By machine learning algorithm, convolutional neural network have been applied to different datasets. We used training set and test set of images of plant seedlings at various stages of grown. The dataset comprises 12 plant species. The goal is to create a classifier capable of determining a plant’s species from a photo. The experimental results validate that the proposed method effectively classifies plant seedlings which are there in dataset. The training set achieved an accuracy of 93% and test set achieved accuracy of 95%. In future works, we plan to detect disease on identified plant species.
Plants remain an important and essential source of food and oxygen for nearly all living organisms on earth. Agriculture is prevailing in some continents like Africa, therefore appropriate automation of the farming procedure would assist in optimizing the crop yield and ensuring the perpetual productivity and sustainability. In accordance with there is a sturdy bond between raised productivity and economic growth. Thus, the application of smart farming techniques in the agricultural sector can empower the development of the economy in many countries. Seedlings quality assessing proved to be a powerful means of prophesying the growth performance and, hence, optimizing the plant production. Seedling classification is the first step to fulfill the seedling quality evaluation.
Furthermore, the invasion of weeds on farmlands leads to decline in the crop yield. Generally, weeds have no valuable beneficial, regarding nutrition, food or medication. However, they grow very quickly as well as they intrusively compete with original crops for space and nutrients. Weeds identification is not an easy process due to the hazy boundaries of the crops, together with the diverse sandy and rocky backgrounds. Thus, there is a need to develop an efficient technique to accurately and certainly detect weeds from beneficial plants.
In order to improve agronomic production and crop quality, farmers should follow precision agriculture.
Precision agriculture is a farm management approach that utilizes information technology and artificial intelligence to guarantee profit maximization, crop yield optimization, and environment preservation. One of the fundamental challenges that face precision agriculture is weed control. Weed control must be achieved earlier as possible after crop germination before weeds begin to compete with crops for nutrition and cause adverse effects. Thus, optimal weed treatment is recommended in the seedling stage. Nevertheless, in this phase, the discrimination between crops and weeds has some limitations;
- a) Inadequate image resolution for distinguishing between exposed soil, crop seedlings and weeds
- b) Resemblance of spectra and appearances between weeds and useful crops in the early stages
- c) Overlapping of the soil background reflectance with the detection process
The plant seedlings dataset contains images of approximately 960 unique plants belonging to 12 species at several growth stages. It comprises annotated RGB images with a physical resolution of roughly 10 pixels per mm.
The application of machine learning techniques for automatic plant seedling classification has become a significant and promising field of research towards improving agriculture outcomes. Deep learning is a specific type of machine learning that has gained substantial interest in various disciplines. The Convolutional Neural Network (CNN) is a deep neural network architecture that is generally used to analyze visual images. Latterly, CNNs have achieved a significant breakthrough in computer vision fields. Additionally, the CNNs proved to have high ability to obtain the efficient features needed for image classification process. Traditional image classification algorithms, handcrafted features are firstly extracted, then a feature selection process is achieved, and finally, a suitable classifier is chosen. However, CNN is proficient in learning various features from images, it covers global and local features, and it uses these features for efficient classification. CNN showed superior performance compared to other image processing techniques. Therefore, in this project, the enforcement of the CNN approach for plant seedling classification is investigated. We are implementing our project using CNN that is Convolutional Neural Network due to its extraordinary features from other existing techniques.
The method proposed by Yang Song et al. (2016) aims to do texture image classification with discriminative neural network. Key area of working is CNN-based features to achieve more accurate classification of texture images, discriminative neural network-based feature transformation (NFT) method for better classification. They come up with enhanced classification performance where CNN-based features (FC-CNN and FVCNN) provides better classification than handcrafted features. Research gap is observed there where evaluation of FV descriptors based on other types of local features that are handcrafted can be done.
The method proposed by Heba A. Elnemr (2019) aims to create a Convolutional Neural Network architecture for plant seedling classification. Key area of working is to use CNN to classify plant seedlings in early growth stage and parameters like accuracy, precision are observed. They come up with accuracy of 94.38%. This system can be used to create IoT system for weed control.
The method proposed by KuoFang Chung et al. (2017) aims to do phylogenetic classification of seed plants of Taiwan. Key area of working is to classify plant species from Taiwan on the basis of biological things. They come up with gymnosperms in 5 families and angiosperms in 210 families. Research gap is observed there where they used dataset consisting of plants only from Taiwan.
The method proposed by Yong Wang et al. (2019) aims to do research on image classification model based on deep convolution neural network. Key area of working is to use depth neural network for maximum interval minimum classification error, M3CE, comparison of SVM, KNN, BP and CNN methods. They come up with accuracy of each methods like CNN,SVM,KNN etc. and they turned out as 99.68%,89.41%,81.25% resp. Deep convolutional neural network as a black box feature extraction model.
Save your time!
We can take care of your essay
- Proper editing and formatting
- Free revision, title page, and bibliography
- Flexible prices and money-back guarantee
The method proposed by Samir S. Yadav et al. (2019) aims to use deep convolutional neural network based medical image classification for disease diagnosis. Key area of working is Deep neural network is capable of classifying images, linear support vector machine classifier with local rotation and orientation free features, transfer learning on two convolutional neural network models: Visual Geometry Group i.e., VGG16 and InceptionV3. Transfer learning deals with retaining specific features and improved performance. They come up with conclusion as CNN-based transfer learning is the best method of all three methods. The capsule network is better than the ORB and SVM classifier. In general, CNN based methods are better than traditional methods. Transfer learning on VGG16 provides best results. Research gap is observed there where they can add visualization at output so that we will get visuals of disease.
The method proposed by Yunus Ozen et al. (2018) aims to detect and classify plant leaf disease based on CNN with LVQ algorithm. Key area of working is to detect diseases like bacterial spot, late blight, septoria leaf spot and yellow curved leaf diseases and necessary action is taken, color information is used for plant leaf disease researches. Leaf disease and classification using CN N and LVQ is done here. Improvement in recognition rate in classification process different filters will be done as far as future scope is considered.
The method proposed by Aydin Kayaa et al. (2018) aims to do analysis of transfer learning for deep neural network based plant classification models. Key area of working is to apply different data sets and use transfer learning to provide important benefits in plant classification. They come up with End-to-end CNN with accuracy 70.79%, LDA and SVM average accuracy reads 95% and 96% resp. for all four plants of dataset. This system can be extended to plant disease identification and weed control.
The method proposed by Jayme Garcia Arnal Barbedo (2013) aims to use digital image processing techniques for detecting, quantifying and classifying plant diseases. Key area of working is to use digital image processing techniques to detect, quantify and classify plant diseases. The selected proposals are divided into three classes as detection, severity quantification, and classification. They come up with detection, severity quantification, and classification for various tools like CNN, SVM etc. Improvement in disease identification, provide pesticides/medicines to cure identified disease are the things on which we can work in future.
The method proposed by Shenghui Yang et al. (2019) aims to use classification method of plug seedlings based on transfer learning. Key area of working is to use classification method for plug seedlings based on transfer learning. Extracting and graying the interest region, average accuracy 100% is achieved for 3 specifications. They come up with accuracy of VGG16 and it turned out 95.50%.
The method proposed by Ruji P. Medina et al. (2018) aims to classify plant seedling images using deep learning. Key area of working is to use image transforms like resize, rotate, flip, scaling and histogram equalization in CNN and data augmentation. They come up with 99.74% accuracy for validation and 99.69% accuracy for testing. This system can be extended to classify herbal plant species from different countries.
The method proposed by Yair Weiss et al. (2019) aims to why deep convolutional networks generalize so poorly to small image transformations? Key area of working is small translations or rescaling’s of the input image can drastically change the network’s prediction. ML tools like CNN, data augmentation and antialiasing are used. They come up with CNN learning invariance from the data augmentation procedure. We can achieve more accuracy though image size is small.
The ability to do so effectively can mean better crop yields and better stewardship of the environment. Huge rise in the number of images that are collected on a daily basis. A large database of images is available worldwide across various platforms. In order to utilize these databases, an effective and robust retrieval and search approach is required. The application of ML techniques for automatic plant seedling classification has become a significant and promising field of research towards improving agriculture outcomes. This process became very laborious and ambiguous due to the rapid increase in the number of images and the diversity of the image contents. Hence, a new method i.e. CNN was devised and is very useful for this purpose.
In area of plant seedling classification, various machine learning algorithms are used. In most of the projects, CNN, SVM, KNN, LVQ etc. algorithms are used and classification is done. Each algorithm achieves different accuracy (in percentage) and has pros and cons. CNN that is convolutional neural network is widely used in image classification as it provides high accuracy as compared to other algorithms. SVM that is state vector machine is another algorithm used in image classification but its accuracy is slightly less than that of CNN. KNN that is K-Nearest Neighbors is another algorithm used in image classification. It is easy to implement. LVQ that is Learning Vector Quantization is another algorithm used in image classification. It is basically prototype based supervised classification algorithm.
CNNs for applications that involve images. Why CNNs are more suitable to process images? Pixels in an image correlate to each other. However, nearby pixels correlate stronger and distant pixels don’t influence much. Local features are important: Local Receptive Fields. Affine transformations: e.g. the class of an image doesn’t change with respect to translation. So we can build a feature detector that can look for a particular feature (e.g. an edge) anywhere in the image plane by moving across. A convolutional layer may have several such filters constituting the depth dimension of the layer. CNN Conceptual Understanding – Inputs: the input layer is a linear arrangement of input elements. This is typically represented as an input vector. CNNS operate on a 3d input ‘volume’. Hidden Layers of the architecture. Neural networks use one or more hidden layers with the units typically performing some nonlinear transformations (e.g. tanh). Convolutional networks use: convolutional layers and pooling layers. Local receptive fields and shared weights for a given filter across the complete input. Strides Filters arranged depth wise: small size. Output layer In a traditional NN the output layer (such as Softmax or Logistic) connect to the linear hidden layer in a fully connected pattern CNNS follow a similar pattern where the ‘volume’ represented by the final internal layer is fully connected to the output layer. CNNs operate on images that have a 2d surface and a depth in terms of RGB colors. Hence they are 3 dimensional. The standard neural network architecture has a linear input layer structure. When we discuss CNN architecture, it helps to decouple in our minds the abstraction CNNs deal with and the practical realization of these principles on a linear layer based neural network architectures. The algorithm here performs in three steps and they are as follows-
- a. Preprocessing- crop parts of image, flip image horizontally, adjust contrast and saturation
- b. Splitting data set
- c. Building CNN
CNN is used to identify, scaling, translation and other forms of images. This project is plant seedlings classification. While doing this project we come to know about various concepts like deep learning, machine learning, CNN i.e. convolution neural network etc. The code had been successfully implemented using CNN in Jupyter and Python IDEs. We also come to know about how plant seedlings are classified using various machine learning tools specially CNN. At output, we got one .csv file which comprises of image file and species which we call it as seedlings classification. In future works, we can detect disease on identified plant seedling. We can classify herbal plants. To increase understanding of details of target object, further research is needed. Another dimension is to perform new experiments when more public datasets become available.
- Yang Song, Qing Li, Dagan Feng and Weidong Cai, “Texture image classification with discriminative neural networks”, Research article on computational visual media Vol. 2, No.4, December 2016, 367-377
- Heba A. Elnemr, “Convolutional neural network architecture for plant seedling classification”, International Journal of Advanced Computer Science and Applications Vol. 10, No. 8, 2019
- Cheng-Tao Lin, Kuo-Fang Chung, “Phylogenetic classification of seed plants of Taiwan”, Lin and Chung Bot Stud (2017) 58:52, DOI 10.1186/s40529-017-0206-6
- Mingyuan Xin, Yong Wang, “Research on image classification model based on deep convolution neural network”, EURASIP Journal on Image and Video Processing, 2019
- Samir S. Yadav, Shivajirao M. Jadhav, “Deep convolutional neural network based medical image classification for disease diagnosis”, Journal of Big Data, 2019
- Yunus Ozen, Adem Tuncer, Melike Sardogan, “Plant leaf disease detection and classification based on CNN with LVQ algorithm”, in: 3rd International Conference on Computer Science and Engineering, 2018
- Aydin Kaya, Ali Seydi Keceli, Cagatay Catal, Hamdi Yalin Yalic, Huseyin Temucin, Bedir Tekinrdogan, “Analysis of transfer learning for deep neural network based plant classification models”, Journal of Computers and Electronics in Agriculture, 2019
- Jayme Garcia Arnal Barbedo, “Digital image processing techniques for detecting, quantifying and classifying plant diseases”, Barbedo SpringerPlus 2013, 2:660
- Zhang Xiao, Yu Tan, Xingxing Liu, Shenghui Yang, “Classification method of plug seedlings based on transfer learning”, Appl. Sci. 2019, 9, 2725; doi:10.3390/app9132725
- Catherine R. Alimboyong, alexander A. Hernandez, Ruji P. Medina, “Classification of plant seedling images using deep learning”, in: Proceedings of TENCON 2018- 2018 IEEE Region 10 Conference
- Aharon Azulay, Yair Weiss, “Why do deep convolutional networks generalize so poorly to small image transformations?”,Journal of Machine Learning Research 20 (2019) 1-25