Assessing the use of object based image analysis and monteith model in sugarcane yield estimation
MetadataShow full item record
Regarding safety, better policy making, timely and appropriate decision making by the manufacturers, government, entrepreneurs and the countries involve in importing sugar, it is important to estimate sugarcane area and yields. This study therefore investigates the use of Object Based Image Analysis and Monteith model in sugarcane yield estimation. One cloud free sentinel 2 satellite image of 21st December, 2018 covering the study area was downloaded from Earth Explorer website and processed in ecogniton developer software. Image objects were generated using a multi-resolution segmentation algorithm and later classification was performed in ecognition developer using nearest neighbour algorithm. The classification accuracy was evaluated using independent ground truth points. The confusion matrix analysis gave the overall classification accuracy of 90.23% with Kappa coefficient of 0.85. After classification, sugarcane acreage and yields were computed and compared with the actual acreage and yield data obtained from Kakira Sugar Limited using percentage difference, bar graphs, and scatter plot graph. For acreage estimation, the total percentage difference was -2.19% and -0.86% for Kabiaza and Karongo sugarcane field blocks respectively. The estimated results showed a strong positive correlation with actual ground sugarcane acreage i.e. r=0.80 and 0.82 for Kabiaza and Karongo division respectively. For yield estimation, the total percentage difference was -1.15% and 9.08% for Kabiaza and Karongo sugarcane field blocks respectively The proposed model yielded results that were close to the actual yields i.e. there was a moderate correlation (r=0.58 and 0.56 for Kabiaza and Karongo division respectively) between the estimated and the actual yields. Further studies should focus on the use of satellite image (s) with sufficient spatial and spectral resolution to separate sugarcane from adjacent competing crops and non-crop vegetative surfaces in order to get better classification results and also determine the exact solar constant for the study area in question to get better yield estimation results.