Outlier detection methods and sensor data fusion for precision agriculture

Abstract

Precision agriculture is a concept regarding the use of technology to increase production yield while preserving and optimizing resources. One of the means to achieve that goal is to use sensors to monitor crops and adjust the cultivation according to its needs. This paper compares different techniques for sensor data fusion and detection and removal of outliers from gathered data to improve sensors accuracy and to identify possible sensor malfunction. As a case study, we monitored an experimental crop of precocious dwarf cashew using soil moisture sensors. Combining generalized ESD method and a weighted outlierrobust Kalman filter generated the best result, leading to more accurate data.

Multilevel data fusion for the internet of things in smart agriculture

“…A simple filtering process using the minimum and maximum values can avoid atypical data (outliers) to cause a wrong decision. Outliers are values that deviate from other readings in a sample, and the readings variability can be one of the causes (Torres et al, 2017). This filtering step can also identify problematic sensors that might require repairs or replacement.…”

Smart & Green: An Internet-of-Things Framework for Smart Irrigation

“…The Fusion service can also treat data created in soil moisture prediction when irrigation management is required for fields without soil moisture sensors. Data fusion consists of processing the data to detect and remove outliers (DRO) and submitting data to a co-operative function (CF), according to a multilevel data fusion architecture [58].…”

“…DRO algorithms are specific to the soil layer monitored. According to [58], the Z-score is more efficient for data series at the most superficial soil layer (e.g., z = 15 cm). In contrast, the Generalized ESD (Extreme Studentized Deviate) algorithm works well with data at depth of z = 45 cm.…”

“…The nRF24L01 is a single-chip radio transceiver for the 2.4-2.5 GHz ISM band. After data transmission, the sensor node hibernates for an hour to save power [58]. We used a Raspberry Pi 3 with the Raspbian OS as the gateway to receive soil moisture data and send them by the CoAP or MQTT client to storage.…”