Village level crop inventory using remote sensing and field survey data

Singh, R. P.; Sridhar, V.; Dadhwal, V. K.; Jaishankar, R.; Neelkanthan, M.; Srivastava, Arun; Bairagi, G. D.; Sharma, Neerja; Raza, Shan E. Ahmed; Sharma, R.; Yadav, Manoj Kumar; Joshi, F K; Purohit, N. L.

doi:10.1007/bf02989996

Cited by 8 publications

(6 citation statements)

References 1 publication

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Despite the large percentage of crowdsourced labels that end up discarded, Plantix has provided more ground data than available previously in similar settings [14,[17][18][19] and comparable in volume to the ground truth collected by Indian state agriculture departments annually (11,469 points nationally in kharif and rabi from 2017-2018) [33]. Therefore, despite the challenges of crowdsourcing, the volume and coverage of datasets like Plantix, along with ever-improving data storage, processing, and smartphone access, make crowdsourcing an increasingly viable and useful alternative to traditional field work.…”

Section: Challenges Of Crowdsourced Labels and Possible Ways Forwardmentioning

confidence: 99%

“…Without publicly-available, government-led field surveys, researchers have either organized their own surveys [14,[17][18][19] or mobilized citizen science efforts [20,21] to obtain ground truth labels in smallholder systems. Most often, these labels are used in conjunction with supervised machine learning methods [14,[17][18][19]21] or classification rules designed by crop experts [22,23]. These approaches have shown that supervised machine learning methods like random forests can achieve some success discerning crop types in smallholder systems, but small field size, high within-crop variation, and low training set size continue to limit map accuracies and the generalizability of models.…”

Section: Introductionmentioning

confidence: 99%

“…Due to the small size of Indian crop fields [28], only the latter can be considered field-level mapping. Maps of wheat [19,29] and sugarcane [30,31] have also been created at local scales, illustrating the potential of remote sensing-based crop mapping across India's diverse agricultural landscape. However, to create national or state-wide multi-crop maps and validate them against field observations requires expansive ground truth datasets that have thus far been unavailable.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Mapping Crop Types in Southeast India with Smartphone Crowdsourcing and Deep Learning

Wang

Tommaso

Faulkner³

et al. 2020

Remote Sensing

View full text Add to dashboard Cite

High resolution satellite imagery and modern machine learning methods hold the potential to fill existing data gaps in where crops are grown around the world at a sub-field level. However, high resolution crop type maps have remained challenging to create in developing regions due to a lack of ground truth labels for model development. In this work, we explore the use of crowdsourced data, Sentinel-2 and DigitalGlobe imagery, and convolutional neural networks (CNNs) for crop type mapping in India. Plantix, a free app that uses image recognition to help farmers diagnose crop diseases, logged 9 million geolocated photos from 2017–2019 in India, 2 million of which are in the states of Andhra Pradesh and Telangana in India. Crop type labels based on farmer-submitted images were added by domain experts and deep CNNs. The resulting dataset of crop type at coordinates is high in volume, but also high in noise due to location inaccuracies, submissions from out-of-field, and labeling errors. We employed a number of steps to clean the dataset, which included training a CNN on very high resolution DigitalGlobe imagery to filter for points that are within a crop field. With this cleaned dataset, we extracted Sentinel time series at each point and trained another CNN to predict the crop type at each pixel. When evaluated on the highest quality subset of crowdsourced data, the CNN distinguishes rice, cotton, and “other” crops with 74% accuracy in a 3-way classification and outperforms a random forest trained on harmonic regression features. Furthermore, model performance remains stable when low quality points are introduced into the training set. Our results illustrate the potential of non-traditional, high-volume/high-noise datasets for crop type mapping, some improvements that neural networks can achieve over random forests, and the robustness of such methods against moderate levels of training set noise. Lastly, we caution that obstacles like the lack of good Sentinel-2 cloud mask, imperfect mobile device location accuracy, and preservation of privacy while improving data access will need to be addressed before crowdsourcing can widely and reliably be used to map crops in smallholder systems.

show abstract

Section: Challenges Of Crowdsourced Labels and Possible Ways Forwardmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Mapping Crop Types in Southeast India with Smartphone Crowdsourcing and Deep Learning

Wang

Tommaso

Faulkner³

et al. 2020

Remote Sensing

View full text Add to dashboard Cite

show abstract

“…Crop biophysical parameters retrieval as input in crop simulation model is a critical need and two most common inputs are leaf area index (LAI) and crop phenology (Sehgal et al, 2005). Small area inventory at village level indicated RS-based area was comparable to field data acquired by National Sample Survey Organization (NSSO) for dominant crops only (Singh et al, 2005). Field-scale LAI retrieval has been demonstrated over wheat in Gujarat (Chaurasia et al, 2006) and recently with Sentinel-2A MSI and Landsat-8 OLI (Dhakar et al, 2021) RS-derived field scale LAI have been assimilated in crop simulation models for making production forecasts (Dhakar et al, 2022).…”

Section: Growing Experience On Field-scale Agricultural Applications ...mentioning

confidence: 99%

Progress and Challenges in Earth Observation Data Applications for Agriculture at Field Scale in India and Small Farm Holdings Regions

Ray

Dadhwal

Navalgund

2022

J Indian Soc Remote Sens

View full text Add to dashboard Cite

The paper traces the evolution of agricultural remote sensing in India through its four phases: initial exploratory and aerial data based (1969)(1970)(1971)(1972)(1973)(1974)(1975)(1976)(1977)(1978)(1979)(1980)(1981)(1982), IRS Utilization Program led (1983( -1995( ), post-IRS-1C launch (1996( -2011 and since the establishment of a dedicated institution for EO-based crop forecasting and other agricultural applications (2012 onward). Published field-scale studies are discussed before introducing the submissions to the special issue of the journal, namely Advanced Geospatial Technologies for Agriculture with Special emphasis on field level Applications. A total of 17 articles grouped under crop discrimination and mapping, crop yield estimation, crop suitability analysis, irrigation management and others (UAV, fodder and field boundary extraction) and 15 of these describe work over India and remaining two have Iran as their study area. These studies are linked to other published work as well as challenges under major issues of data access, use of UAV, use of smartphones, data processing and crop simulation models. Taken together, these articles report substantial improvements in capabilities of field-scale monitoring for developing new class of agricultural applications in India as well as other regions in the world with small farm holdings.

show abstract

“…The crops that were discriminated were paddy, maize, and sugarcane [17]. Since 2000 to 2005 Southwestern Brazilian Amazon used MODIS time series data and applied wavelet transformation to determine the growth of row crops and raising the number of crops grown in the area, they got an overall accuracy of 94% [22].…”

Section: Introductionmentioning

confidence: 99%

Rule base Knowledge and Fuzzy Approach for Classification of Specific Crop and Acreage Estimation

Thorat¹,

Kale²,

Mehrotra³

et al. 2017

IJCA

View full text Add to dashboard Cite

Estimation of specific crop and acreage plays a vital role in the field of crop planning, monitoring, crop condition, yield forecasting and acreage estimation. There have been several studies conducted to classify the crops at continental to the regional level, but still, work is needed to map small area covered by different crops using Remote Sensing technology. The main objective of the present study is to explore whether the Fuzzy classifier can improve the accuracy of crop classification as compared to other traditional Classifiers, such as Maximum likelihood, Mahalanobis etc. The attempt has been done to classify different crops at a smaller scale. The Landsat time series 8 band OLI data was used to investigate multiple crop phenomena. Two scenes were acquired in Kharif seasons (September 28 and October 30, 2014). Three indices such as NDVI, SAVI, and RVI, were used to know vegetation condition. The Spectral signatures generated from data for the residues of Sugarcane and Maize based on prior knowledge of the field work. Four techniques based on Maximum Likelihood, Mahalanobis Classifier, Knowledge classifier and fuzzy classification techniques were used to extract the crops information based on the signatures. The resulting overall classification accuracy was calculated using stratified random sampling method. The corresponding performance efficiency of these four methods was found to be 84%, 85%, 87% and 90.67%, respectively, indicating the fuzzy method to be the most efficient as compared with other classification techniques.

show abstract

Village level crop inventory using remote sensing and field survey data

Cited by 8 publications

References 1 publication

Mapping Crop Types in Southeast India with Smartphone Crowdsourcing and Deep Learning

Mapping Crop Types in Southeast India with Smartphone Crowdsourcing and Deep Learning

Progress and Challenges in Earth Observation Data Applications for Agriculture at Field Scale in India and Small Farm Holdings Regions

Rule base Knowledge and Fuzzy Approach for Classification of Specific Crop and Acreage Estimation

Contact Info

Product

Resources

About