Classification of paddy growth age detection through aerial photograph drone devices using support vector machine and histogram methods, case study of Merauke regency

Chia sẻ: Nguyễn Thảo | Ngày: | Loại File: PDF | Số trang:10

Thêm vào BST

Báo xấu

14
lượt xem 1
download

Download Vui lòng tải xuống để xem tài liệu đầy đủ

In this paper we present an approach to estimate the age of paddy in drone images using the Support Vector Machines - SVM and Histogram method.

Chủ đề:

Bình luận(0) Đăng nhập để gửi bình luận!

Lưu

Nội dung Text: Classification of paddy growth age detection through aerial photograph drone devices using support vector machine and histogram methods, case study of Merauke regency

International Journal of Mechanical Engineering and Technology (IJMET) Volume 10, Issue 03, March 2019, pp. 1850-1859. Article ID: IJMET_10_03_187 Available online at http://www.iaeme.com/ijmet/issues.asp?JType=IJMET&VType=10&IType=3 ISSN Print: 0976-6340 and ISSN Online: 0976-6359 © IAEME Publication Scopus Indexed CLASSIFICATION OF PADDY GROWTH AGE DETECTION THROUGH AERIAL PHOTOGRAPH DRONE DEVICES USING SUPPORT VECTOR MACHINE AND HISTOGRAM METHODS, CASE STUDY OF MERAUKE REGENCY Marsujitullah, Fransiskus X. Manggau and Rachmat Informatics Engineering, Universitas Musamus, Merauke, Indonesia ABSTRACT Farming is one of the spearheads of national development which has an important role, especially Merauke Regency which is planned as an area of national food self- sufficiency in the field of agribusiness. Agriculture in Indonesia has a lot of food land that is widely spread and various types of paddy fields from several types of food management especially in agriculture, however there is no system that visualizes the progress of food crop growth in particular areas by looking at the condition of the land in an approach visual. The estimated age of paddy growth is aimed at managing and monitoring paddy plants as information needs in assisting the government, especially in monitoring the area planted by utilizing image images taken through aerial photographs using Drone devices. In this paper we present an approach to estimate the age of paddy in drone images using the Support Vector Machines - SVM and Histogram method. SVM is a learning machine method that works on the principle of Structural Risk Minimization (SRM) with the aim of finding the best hyperplane that separates two classes in input space. Input data are images from drone devices to support vector machines in their ability to find the best hyperplane that separates two classes in the feature space supported by the SRM strategy. Histograms in graphical form that describe the spread of pixel intensity values of an image. With this research, it can be known the age of paddy plants through the histogram value taken on the image by the drone device, so that the growth phase parameters from one week to the harvest can be known with 89 percent accuracy. Keywords: Support Vector Machines, Histogram, Image classification, Structural Risk Minimization Cite this Article Marsujitullah, Fransiskus X. Manggau and Rachmat, Classification of Paddy Growth Age Detection Through Aerial Photograph Drone Devices Using Support Vector Machine and Histogram Methods, Case Study of Merauke Regency, International Journal of Mechanical Engineering and Technology, 10(3), 2019, pp. 1850-1859. http://www.iaeme.com/IJMET/issues.asp?JType=IJMET&VType=10&IType=3 http://www.iaeme.com/IJMET/index.asp 1850 editor@iaeme.com
Students’ Perceptions towards the Grammar Teaching at English Literature Department of Musamus University 1. INTRODUCTION The purpose of remote sensing is to analyze or measure the physical number of drivers without direct physical contact, the benefits obtained in using remote sensing, especially in using a drone device, are observed in a large area, in terms of financing more affordable compared to land surveys . In its use, drone devices can be used anywhere on agricultural land, considering that Indonesia is not only well-known as an archipelagic country, it is also known by the world community as an agricultural country, where most of its land area is still used for agricultural and plantation purposes. As a country that has been involved in agriculture and plantations for a long time, of course it has often faced various inhibiting factors which can reduce the level of agricultural productivity. Various steps are taken to map agricultural productivity, from simple methods to the use of advanced technologies that exist today. One of the main applications for remote sensing is for agricultural monitoring or management. For example, long distance sensing techniques are used to calculate the number of oil palm trees [1]. In [2], satellite imagery is used to estimate the biomass of secondary forest above land in Brazil. By using remote sensing, monitoring large quantities of agriculture is possible and cost effective. In [3], multispectral satellite imagery, FORMOSAT-2, is used to map cultivation areas and monitor crop status on a regional scale. Furthermore far sensing can also be used to detect plantation areas [4] and crop yields prediction [5]. By way of implanting sensors in agricultural areas as in research [6,7] then processing the sensing data from different places. For each data retrieval method there are advantages and disadvantages. In taking data directly in the field, a lot of time will be wasted because each farm must be visited one by one to get the expected data. However, this will be very different if you use the remote sensing method, where the desired data can be obtained on such a wide scale and in a relatively short time. What is an obstacle in remote sensing is that the earth's atmosphere is not always clean of clouds. Though clouds or other objects in the atmosphere can interfere with or make satellites unable to record events that occur on the surface of the earth. In the data collection method that utilizes sensors planted in each agricultural area the results can be far better when compared to the first and second data collection methods because the tools have been installed on the site so that the accuracy of the reading can be better and do not need to check the farm area one by one so you can save time collecting data. However, unfortunately, the third method requires a considerable investment cost because sensors must be installed in each agricultural area so that in this study, the authors only used the first and second methods. The first method, which is collecting data directly in the field, will be used to cover up the weaknesses of the second method, which is collecting data using remote control, when the earth's atmosphere is covered with clouds so that remote sensing is not possible. In the use of satellite imagery the weaknesses are almost the same as the weaknesses found in the remote sensing method in general, which is very dependent on weather such as rain, clouds and fog caused by the operation of space where shooting still cannot penetrate the clouds. In addition, the maximum spatial resolution that can only reach 15 meters makes it unable to properly detect the types of plants on agricultural land with such a small planting area. The method we do has the same method of retrieving data as the method of taking pictures remotely, that is by using imagery. The image used in this method comes from drones with the height of the rules of flight drones in Indonesia, which is a maximum of 150 meters. From this height one image data can cover ± 3 Ha. http://www.iaeme.com/IJMET/index.asp 1851 editor@iaeme.com
Classification of Paddy Growth Age Detection Through Aerial Photograph Drone Devices Using Support Vector Machine and Histogram Methods, Case Study of Merauke Regency 2. DEVELOPMENT OF PADDY AGE GROWTH There are a number of developments in the growth that paddy plants go through from the beginning of planting to entering the harvest period with varying spectral signature characteristics when viewed using Landsat TM satellite imagery among them. 1) The initial development of growing paddy, where paddy fields are dominated by water due to flooding. In Landsat TM images with the true color composite (TCC) color composition, the paddy fields will appear blue. 2) The development of the second stage of vegetative growth, marked by the thickening of the leaves of paddy plants that cover all paddy fields. At this stage, land cover is dominated by green. This green color will appear green in the image. 3) Development of the third stage of generative growth, in which paddy fields which were originally dominated by green leaves will be replaced with paddy pale yellow in color on TCC. 4) Development of stage four where land becomes paddy for a certain period of time. In this condition the wetland will appear reddish brown in the TCC color composition. The development of paddy crops during harvest can be estimated using Landsat satellite images if referring to the average age of paddy ranging from 110-120 days. This can be done by first monitoring the start of planting, which is a change from the fallow phase (land preparation stage) to the water phase (tillage / flooding) or by monitoring changes in the stage of paddy from the water stage to the vegetative stage. Figure 1. Development of age for paddy growth. (a). initial stage of growth, (b). Vegetative growth stage, (c). Generative growth stage, (d). Ready for harvest. 3. HISTOGRAM AND SVM DEVELOPMENT OF PADDY GROWTH 3.1. Histogram The use of digital images is increasing because of the advantages possessed by digital images, including the ease of getting pictures, reproducing images, processing images and others. But not all digital images have a visual display that satisfies the human eye. Dissatisfaction can arise due to noise, the lighting quality in digital images that are too dark or too bright. So that a method is needed to improve the quality of the digital image. To improve image quality in terms of color contrast, we can give treatment to the histogram. The treatment referred to in this article is an equalization histogram on grayscale images. Image histogram is said to be good if it is able to involve all possible levels or levels at the gray level. Of course the aim is to be able to display details on the image so that it is easy to observe [8]. http://www.iaeme.com/IJMET/index.asp 1852 editor@iaeme.com
Students’ Perceptions towards the Grammar Teaching at English Literature Department of Musamus University Many histograms are applied in some cases. In the study [9] histograms were used to identify age through the face. The histogram technique is also used in [10] for health. To determine the growth phase needs of rice plants, leaf color is the easiest indicator. Giving the amount and at the right time can provide an increase in the efficiency of the actual absorption of the plants, so that the yield is as expected. The use of leaf color chart by equating the color of rice leaves with a color scale composed of green series, ranging from yellowish green to dark green accompanied by very important parameters to facilitate the classification of the rice. Leaf Color Chart (LCC) is a leaf color level standard issued by the International Paddy Research Institute (IRRI). LCC is usually used to determine the nitrogen content of a plant so that later it can be known when the fertilization and harvest time is right. The use of cameras on drone drones in leaf shooting will help farmers to determine the color level of plants automatically, in this case helping the government in obtaining agricultural information based on the LCC. In addition, farmers usually use the LCC manually by comparing the color of the plant leaves with each color level found on the LCC. Determination of LCC level can be done automatically by utilizing the image so that farmers are expected to know the information on the image of the leaf located at what level in the leaf color chart. 3.2. Support Vector Machine - SVM Support Vector Machine (SVM) was first proposed for classification problems. SVM is used to analyze voltage [11]. In [12] also SVM is used for spectral-spatial. This is a learning technique for supervised non-parametric statistics. Therefore, the main advantage is that data distribution does not need to be known in priority, whereas other statistical techniques, for example, maximum likelihood estimates usually assume that data distribution is known as a priority [13-18]. To explain the concept of supporting vector machines, a classifcation problem of two linear classes is used, see Figure 2. The purpose of vector support machine techniques is to find a hyperplane separating data into many classes, which are two classes in this case. A hyperplane like this is called a decision boundar or an SVM hyperplane. To get a unique hyplane or optimal separation, an obstacle is that there is no data point in the hyperplane margin. Data points on the margin are called vector support. In other words, supporting vectors are used to defect the maximum margin hyperplane. If data is not distributed linearly, using a hyperplane cannot separate data into many classes efficiently. To handle the distribution of non-linear data, data is projected into higher dimensional spaces so that data points are distributed linearly in new spaces. By using the right projection function, products in higher dimensional space can be calculated in the original space without mapping data points into feature spaces that may have infinite dimensions through the use of the kerel function. Figure. 2. Illustration of supporting vector machine concepts. http://www.iaeme.com/IJMET/index.asp 1853 editor@iaeme.com
Classification of Paddy Growth Age Detection Through Aerial Photograph Drone Devices Using Support Vector Machine and Histogram Methods, Case Study of Merauke Regency 3.3. Method”oneagainstall” Using this method, a binary SVM model is constructed (k is the number of classes). Each i classification model is trained by using all data, to find solutions to problems. For example, there are classification problems with 4 classes. For training, 4 binary SVM are used as in table 1. 𝑚𝑖𝑛 1 𝑇 𝑤 𝑖 ,𝑏𝑖 ,𝜉 𝑗 2 (𝑤 𝑖 ) 𝑤 𝑖 + 𝐶 ∑𝑡 𝜉𝑡𝑖 (1) 𝑇 𝑠. 𝑡 (𝑤 𝑖 ) ∅ (𝑥𝑡 ) + 𝑏 𝑖 ≥ 1 − 𝜉𝑡𝑖 → 𝑦𝑡 = 𝑖, (2) 𝑇 (𝑤 𝑖 ) ∅ (𝑥𝑡 ) + 𝑏 𝑖 ≥ 1 − 𝜉𝑡𝑖 → 𝑦𝑡 ≠ 𝑖, (3) 𝜉𝑡𝑖 ≥ 0 (4) 4. CASE STUDY AND RESULTS As mentioned earlier, the purpose of this study was to detect the growth phase of rice by dividing the rice group from 0 weeks to post-harvest. So that the information needs of an area can be seen from the image taking of rice by a drone device which then calculates the histogram value and determines the age group of rice using Support Vector Machine - SVM, with accurate calculations carried out histogram of each image, helps SVM in determining the age classification of rice, so that the government can control the condition of an area's rice fields easily. The monitored variable refers to several RGB color variables from the image of rice from each image capture using a Drone device. Some measurements of the accuracy of the data obtained do not produce 100% results, considering the color variables of each image are influenced by sunlight and the height of shooting using a Drone device, but the accuracy obtained is 89.00%. The following is the process in the design of the running system that is made using training data for rice ages 3 to 4 months, and 3-4 months test data, and predicts other test data with system errors using test data for rice ages 0-3 weeks. Figure 3. Data Training Load Data Retrieval of image data to be stored http://www.iaeme.com/IJMET/index.asp 1854 editor@iaeme.com
Students’ Perceptions towards the Grammar Teaching at English Literature Department of Musamus University Figure 4. Data Training Load Data Figure 5. Data Successful in Training Figure 6. Data Prediction Image of Paddy Age Appropriate Figure 7. Paddy Age Image Prediction Does Not Match The running system design process describes taking image data to calculate histograms and stored in a database until the prediction of rice age testing using test data. The first step that is done is taking pictures of training data and then calculating the histogram and save the image data, in (figure 3). After the data is stored, the data will be tested using the test data that has been provided, the test shooting stage is carried out, then count the histogram, and load the data (figure 4) which is continued in the training phase using SVM (figure 5). Then the image data prediction process will be carried out whether it will be in accordance with the test data that has not been trained by the system, if true, the data will display the age of the rice according to the http://www.iaeme.com/IJMET/index.asp 1855 editor@iaeme.com
Classification of Paddy Growth Age Detection Through Aerial Photograph Drone Devices Using Support Vector Machine and Histogram Methods, Case Study of Merauke Regency image data that has been properly trained (figure 6), but if it does not work the system will display the results with an incorrect command (figure 7). Of the 200 data available, 140 data were selected for training to represent each age of rice determined and will then be stored in the database, 60 other data provided for testing by loading data, SVM training on image data, then histogram values and data predictions will be calculated picture. The prediction of the accuracy of the data that has been tested will be determined by the correct information rating divided by the total test data at 100%. 𝑇𝑟𝑢𝑒 𝐴𝑐𝑐𝑢𝑟𝑎𝑐𝑦 = 𝑇𝑜𝑡𝑎𝑙 × 100 % (5) For the next experiment, it can be seen in the test results table with 60 data tested, with each data, namely the initial harvest period of 15 data, the vegetative period of 15 data, the period of 15 data and the harvest period of 15 data. Table 1. Results of experimental test data Paddy Age Group Explanation No Y Y True False (Original) (Prediction) 1 0 – 3 Weeks 0 – 3 Weeks √ - 2 0 – 3 Weeks 0 – 3 Weeks √ - 3 0 – 3 Weeks 0 – 3 Weeks √ - 4 0 – 3 Weeks 0 – 3 Weeks √ - 5 0 – 3 Weeks 0 – 3 Weeks √ - 6 0 – 3 Weeks 0 – 3 Weeks √ - 7 0 – 3 Weeks 0 – 3 Weeks √ - 8 0 – 3 Weeks 0 – 3 Weeks √ - 9 0 – 3 Weeks 0 – 3 Weeks √ - 10 0 – 3 Weeks 0 – 3 Weeks √ - 11 0 – 3 Weeks 0 – 3 Weeks × 12 0 – 3 Weeks 0 – 3 Weeks √ - 13 0 – 3 Weeks 0 – 3 Weeks √ - 14 0 – 3 Weeks 0 – 3 Weeks × 15 0 – 3 Weeks 0 – 3 Weeks √ - 16 3 Weeks – 2 Months 3 Weeks – 2 Months √ - 17 3 Weeks – 2 Months 3 Weeks – 2 Months √ - 18 3 Weeks – 2 Months 3 Weeks – 2 Months √ - 19 3 Weeks – 2 Months 3 Weeks – 2 Months √ - 20 3 Weeks – 2 Months 3 Weeks – 2 Months × 21 3 Weeks – 2 Months 3 Weeks – 2 Months √ - 22 3 Weeks – 2 Months 3 Weeks – 2 Months √ - 23 3 Weeks – 2 Months 3 Weeks – 2 Months √ - 24 3 Weeks – 2 Months 3 Weeks – 2 Months √ - 25 3 Weeks – 2 Months 3 Weeks – 2 Months √ - 26 3 Weeks – 2 Months 3 Weeks – 2 Months √ - 27 3 Weeks – 2 Months 3 Weeks – 2 Months √ - 28 3 Weeks – 2 Months 3 Weeks – 2 Months √ - 29 3 Weeks – 2 Months 3 Weeks – 2 Months × http://www.iaeme.com/IJMET/index.asp 1856 editor@iaeme.com
Students’ Perceptions towards the Grammar Teaching at English Literature Department of Musamus University 30 3 Weeks – 2 Months 3 Weeks – 2 Months √ - 31 2Months – 3Months 2Months – 3Months √ - 32 2Months – 3Months 2Months – 3Months √ - 33 2Months – 3Months 2Months – 3Months √ - 34 2Months – 3Months 2Months – 3Months √ - 35 2Months – 3Months 2Months – 3Months √ - 36 2Months – 3Months 2Months – 3Months × 37 2Months – 3Months 2Months – 3Months √ - 38 2Months – 3Months 2Months – 3Months √ - 39 2Months – 3Months 2Months – 3Months √ - 40 2Months – 3Months 2Months – 3Months × 41 2Months – 3Months 2Months – 3Months √ - 42 2Months – 3Months 2Months – 3Months √ - 43 2Months – 3Months 2Months – 3Months √ - 44 2Months – 3Months 2Months – 3Months √ - 45 2Months – 3Months 2Months – 3Months √ - 46 3Months – 4Months 3Months – 4Months √ - 47 3Months – 4Months 3Months – 4Months √ - 48 3Months – 4Months 3Months – 4Months √ - 49 3Months – 4Months 3Months – 4Months √ - 50 3Months – 4Months 3Months – 4Months × 51 3Months – 4Months 3Months – 4Months √ - 52 3Months – 4Months 3Months – 4Months √ - 53 3Months – 4Months 3Months – 4Months √ - 54 3Months – 4Months 3Months – 4Months √ - 55 3Months – 4Months 3Months – 4Months × 56 3Months – 4Months 3Months – 4Months √ - 57 3Months – 4Months 3Months – 4Months √ - 58 3Months – 4Months 3Months – 4Months √ - 59 3Months – 4Months 3Months – 4Months √ - 60 3Months – 4Months 3Months – 4Months √ - 5. CONCLUSION From the results of experiments using histogram and SVM calculations to determine the development of rice growth from the four planting periods can be concluded: 1. The results obtained from the calculation of image data for rice harvested through a Drone device are 89.00%, with errors obtained reaching 11.00%, indicating that the proposed methodology was successfully used. 2. Research carried out using 200 data as a whole, and there were 22 data that were wrong, with the correctness of data 178 data. 3. Using the Histogram methodology and Support Vector Machine, operators can quickly find out the identification of developments in rice age growth by utilizing images from drone devices to assist information needs for government through agricultural extension workers. http://www.iaeme.com/IJMET/index.asp 1857 editor@iaeme.com
Classification of Paddy Growth Age Detection Through Aerial Photograph Drone Devices Using Support Vector Machine and Histogram Methods, Case Study of Merauke Regency REFERENCES [1] P. Srestasathiern and P. Rakwatin, "Oil palm tree detection with high resolution multi- spectral satellite imagery," Remote Sensing, vol. 6, no. 10, pp. 9749-9774, 2014. [2] M. K. Steininger, "Satellite estimation of tropical secondary forest above-ground biomass: data from brazil and bolivia," International Journal Of Remote Sensing, vol. 21, no. 6-7, pp. 1139-1157, 2000. [3] Q. Zhao, Y. I. Lenz-Wiedemann, F. Yuan, R. Jiang, Y. Miao, F. Zhang, and G. Bareth, "Investigating within-field variability of paddy from high resolution satellite imagery in qixing farm county, northeast china," ISPRS International Journal of Geo-information, vol. 4, pp. 236-261, 2015. [4] M. K. Mosleh and Q. K. Hassan, "Development of a remote sensing based boro paddy mapping system," Remote sensing, vol. 6, pp. 1938-1953, 2014. [5] M. K. Mosleh, Q. K. Hassan, and E. H. Chowdhury, "Application of remote sensors in mapping paddy area and forecasting its production: A review," Sensors, vol. 15, pp. 769- 791, 2015. [6] Scarlett Liu, Samuel Marden, Mark Whitty. 2013. Towards Automated Yield Estimation in Viticulture. Proceedings of Australasian Conference on Robotics and Automation, University of New South Wales, Sydney Australia. [7] Qi Wang, Stephen Nuske, Marcel Bergerman, Sanjiv Singh. 2013. Automated Crop Yield Estimation for Apple Orchards. J.P. Desai et al. (Eds): Experimental Robotics, STAR 88, pp. 745 - 758. [8] Nazaruddin Ahmad, Arifyanto Hadinegoro,”Metode Histogram Equalization Untuk Perbaikan Citra Digital”, Seminar Nasional Teknologi Informasi & Komunikasi Terapan 2012. [9] A. Deppa and T. Sasipraba, “Age estimation in facial images using histogram equalization”, Eighth International Conference on Advanced Computing (ICoAC), 2016. [10] Olivassé Nasario-Junior1, Paulo R. Benchimol-Barbosa1,2, Jurandir Nadal1, “Beat-to-beat T-peak T-end Interval Duration Variability Assessed by RR-Interval Histogram Analysis in Health Sedentary and Athlete”, Computing in Cardiology 2017; VOL 44. [11] Antonin Dem. Iven Mareels and Sajeeb Saha, “Assessment of Voltage Stability Risks under Stochastic Net Loads using Scalable SVM Classiﬁcation”, Australasian Universities Power Engineering Conference (AUPEC), 2017 . [12] Rafika Ben Salem and Mohamed Ali Hamdi, “Fuzzy C-mean for unsupervised spectral- spatial SVM classification of hyperspectral images”, ACS 14th International Conference. [13] Gerzon Jokomen Maulany, Fransiskus Xaverius Manggau, Jayadi, Richard Semuel Waremra and Casimirus Andy Fenanlampir, Radiation Detection of Alfa, Beta, and Gamma Rays with Geiger Muller Detector, International Journal of Mechanical Engineering and Technology, 9(11), 2018, pp. 21–27. [14] Nasra Pratama Putra and Gerzon Jokomen Maulany, Classification System for Student Study Duration on Department of Information Systems at Musamus University, Using Id3, International Journal of Mechanical Engineering and Technology, 9(12), 2018, pp. 878– 885. [15] Stanly Hence Dolfi Loppies and Gerzon Jokomen Maulany, Geographic Information System Location of Pre-Prosperous Family Housing of Merauke District, International Journal of Mechanical Engineering and Technology, 9(12), 2018, pp. 177–183. [16] Mangkoedihardjo, S. and Samudro, G. Research strategy on kenaf for phytoremediation of organic matter and metals polluted soil. Advances in Environmental Biology, 8(17), 2014, pp. 64-67. [17] Utomo, A.A. and Mangkoedihardjo, S. Preliminary Assessment of Mixed Plants for Phytoremediation of Chromium Contaminated Soil. Current World Environment, 13(Special Issue 1), 2018, pp. 22-24. http://www.iaeme.com/IJMET/index.asp 1858 editor@iaeme.com
Students’ Perceptions towards the Grammar Teaching at English Literature Department of Musamus University [18] Zaman, B; Purwanto, Mangkoedihardjo, S. Reversible Anaerob-Evapotranspiration Process for Removal of High Strength Ammonium in Leachate from Tropical Landfill. Advanced Science Letters, 23(3), 2017, pp. 2586-2588. http://www.iaeme.com/IJMET/index.asp 1859 editor@iaeme.com