Systems and methods for detecting disguised non-performing assets (NPAs) and predicting potential NPAs in agriculture

A system, comprising:

• a memory storing instructions;
• one or more communication interfaces;
• and
• one or more hardware processors coupled to the memory via the one or more communication interfaces, wherein the one or more hardware processors are configured by the instructions to:
• obtain one or more satellite images corresponding to one or more crops;
• identify a type of the one or more crops cultivated on a specific land based on the one or more satellite images using one or more machine learning (ML) or deep learning (DL) algorithms, wherein the one or more ML or DL algorithms are trained with a dataset of the one or more satellite images where the one or more crops in the dataset are identified and annotated, wherein specific pixels in the dataset of the one or more satellite images are identified to train the one or more ML or DL algorithms to identify the type of the one or more crops cultivated in the specific pixels, and wherein if the type of the one or more crops identified from the one or more satellite images and provided by a farmer to a bank is same then value=1 is assigned, else value=2 is assigned;
• estimate temporal health of each of the one or more crops using the one or more obtained satellite images based on the identified type, wherein for each time instance, when the one or more satellite images are available, number of indices including Normalized Difference Vegetation Index (NDVI) are estimated from the one or more satellite images and the temporal health of each of the one or more crops has a score between 0 and 1 where 0 indicates poor health and 1 indicates ideal health, and wherein the score is estimated using a field boundary available for a target plot, the NDVI for the target plot is obtained and compared with an average of the NDVI for the same crop from a surrounding region;
• obtain information pertaining to agroclimatic (AC) zone, and natural calamity from one or more sources;
• estimate crop losses for each of the one or more crops based on value associated with occurrence of the natural calamity;
• estimate yield for each of the one or more crops based on at least one or more satellite images, and weather data being obtained;
• compute a first score as a function of the type of the one or more crops, the temporal health, the agroclimatic (AC) zone, the natural calamity, the crop losses, the estimated yield, and associated market returns being obtained;
• detect the one or more crops as one of (i) a first non-performing asset type, or (ii) a second non-performing asset type based on the computed first score, wherein the first non-performing asset type is a disguised Non-Performing Asset (NPA) and the second non-performing asset type is a non-disguised NPA, and wherein if the computed first score is more than 0.5, then assign score=1 which indicates the non-disguised NPA and if the computed first score is less than 0.5, then assign score=0 which indicates the disguised NPA;
• and
• periodically performing, via the one or more hardware processors, the steps of:
• monitoring growth of the one or more crops based on at least one of the one or more satellite images and sensory information obtained from one or more sensors;
• monitoring crop protocol associated with the one or more crops, wherein the crop protocol comprises dynamically generated instructions for cultivating the one or more crops, wherein the instructions are empirically determined and generated based on image processing of the one or more satellite images, historical pattern learnt over a period of time, wherein the instructions are in the form of advice wherein with each of monitored parameters, personalized advice are provided to farmers including (i) crop health wherein based on the crop health of fields, which portions of the fields (locations/hotspots) needs monitoring, (ii) pest/disease detection wherein based on detected pest/disease and its severity, which chemicals/pesticides to use along with quantity, and (iii) soil moisture sensors based on actual soil moisture measurement, suggestions on when and how to irrigate, wherein based on number of advices followed by the farmers an advice score is generated, wherein the crop protocol include feedback and the feedback include (a) how the activities are performed, (b) based on satellite image data, the system identifies if the farmer has taken certain measures to minimize/remove hotspots, (c) using hyperspectral data the system identifies if the advice on pest/disease has been followed or any further feedback is to be provided and (d) identification by the system if irrigation was provided at right time and with required quantity, and
• wherein the steps of monitoring the growth of the one or more crops and monitoring the crop protocol are performed and updated at periodic time intervals.

One or more non-transitory machine-readable information storage mediums comprising one or more instructions which when executed by one or more hardware processors cause:

• obtaining one or more satellite images corresponding to one or more crops;
• identifying a type of the one or more crops cultivated on a specific land based on the one or more satellite images using one or more machine learning (ML) or deep learning (DL) algorithms, wherein the one or more ML or DL algorithms are trained with a dataset of the one or more satellite images where the one or more crops in the dataset are identified and annotated, wherein specific pixels in the dataset of the one or more satellite images are identified to train the one or more ML or DL algorithms to identify the type of the one or more crops cultivated in the specific pixels, and wherein if the type of the one or more crops identified from the one or more satellite images and provided by a farmer to a bank is same then value=1 is assigned, else value=2 is assigned;
• estimating temporal health of each of the one or more crops using the one or more obtained satellite images based on the identified type, wherein for each time instance, when the one or more satellite images are available, number of indices including Normalized Difference Vegetation Index (NDVI) are estimated from the one or more satellite images and the temporal health of each of the one or more crops has a score between 0 and 1 where 0 indicates poor health and 1 indicates ideal health, and wherein the score is estimated using a field boundary available for a target plot, the NDVI for the target plot is obtained and compared with an average of the NDVI for the same crop from a surrounding region;
• obtaining information pertaining to agroclimatic (AC) zone, and natural calamity from one or more sources;
• estimating crop losses for each of the one or more crops based on value associated with occurrence of the natural calamity;
• estimating yield for each of the one or more crops based on at least one or more satellite images, and weather data being obtained;
• computing a first score as a function of the type of the one or more crops, the temporal health, the agroclimatic (AC) zone, the natural calamity, the crop losses, the estimated yield, and associated market returns being obtained;
• detecting the one or more crops as one of (i) a first non-performing asset type, or (ii) a second non-performing asset type based on the computed first score, wherein the first non-performing asset type is a disguised Non-Performing Asset (NPA) and the second non-performing asset type is a non-disguised NPA, and wherein if the computed first score is more than 0.5, then assign score=1 which indicates the non-disguised NPA and if the computed first score is less than 0.5, then assign score=0 which indicates the disguised NPA;
• and
• periodically performing, via the one or more hardware processors, the steps of:
• monitoring growth of the one or more crops based on at least one of the one or more satellite images and sensory information obtained from one or more sensors;
• monitoring crop protocol associated with the one or more crops, wherein the crop protocol comprises dynamically generated instructions for cultivating the one or more crops, wherein the instructions are empirically determined and generated based on image processing of the one or more satellite images, historical pattern learnt over a period of time, wherein the instructions are in the form of advice wherein with each of monitored parameters, personalized advice are provided to farmers including (i) crop health wherein based on the crop health of fields, which portions of the fields (locations/hotspots) needs monitoring, (ii) pest/disease detection wherein based on detected pest/disease and its severity, which chemicals/pesticides to use along with quantity, and (iii) soil moisture sensors based on actual soil moisture measurement, suggestions on when and how to irrigate, wherein based on number of advices followed by the farmers an advice score is generated, wherein the crop protocol include feedback and the feedback include (a) how the activities are performed, (b) based on satellite image data, a system identifies if the farmer has taken certain measures to minimize/remove hotspots, (c) using hyperspectral data the system identifies if the advice on pest/disease has been followed or any further feedback is to be provided and (d) identification by the system if irrigation was provided at right time and with required quantity, and
• wherein the steps of monitoring the growth of the one or more crops and monitoring the crop protocol are performed and updated at periodic time intervals.