Agricultural activity assessment and fine crop classification of plots based on time-series remote sensing imagery

With the development of remote sensing technology, agricultural monitoring and management are gradually moving towards intelligence and precision. Using time-series remote sensing imagery to determine agricultural activities and classify crops in specific plots has become an important tool in modern agriculture.

Overview of Time-Series Remote Sensing Imagery

Time-series remote sensing imagery refers to multi-temporal imagery data of the same area acquired at different times. These images can come from various platforms such as satellites and drones. By analyzing this data, phenological information in the time dimension and crop distribution information in the spatial dimension can be obtained, thereby revealing the dynamic changes in land cover and crop growth.

Agricultural activity judgment

Assessing agricultural activities primarily involves identifying planting areas, classifying crop types, and monitoring growth status. Time-series remote sensing imagery can achieve the following functions:

Planting area identification:

By analyzing vegetation index changes in multi-temporal images, the distribution of cultivated land in different seasons can be identified.

Machine learning algorithms are used to classify images and accurately delineate planting areas.

Crop type classification:

By utilizing the changes in spectral characteristics of different crops during their growth cycle, a classification model is established to identify crop types.

Combining time series data can improve the accuracy and stability of classification.

Growth monitoring:

By analyzing the temporal changes in vegetation indices, we can monitor crop growth and health. This allows for the timely detection of anomalies within the plot, such as pests and diseases, droughts, and floods, providing early warning information.

Application Examples

1. Visualization of time series data and extraction of phenological periods

△Figure 1. Time series images of the plot and extraction of phenological periods

We collected 2022-2023 quarterly time-series imagery data from a demonstration plot in Chang'an District, Xi'an. After cloud-based filtering and masking, we ultimately selected 18 images. We normalized the crop growth using NDVI to compare differences in crop growth at different times.

As shown in the figure above, by observing the changes in vegetation index over time, key phenological periods can be identified, thereby enabling the prediction and assessment of crop and agricultural activities. Combined with spatial distribution information, crop distribution maps classified by season can be drawn.

2. Based on emerging spatiotemporal hotspot analysis and crop classification of agricultural activities

△Figure 2 Crop distribution map based on agricultural activities

Time-series imagery is converted into vector points or multidimensional raster data, and a spatiotemporal cube is constructed. Based on the cube, hotspot analysis is applied to each time layer according to fixed spatial neighborhood relationships and time steps. The hotspot analysis results are then statistically analyzed for each grid, ultimately yielding a classification map containing both spatial and temporal information. Finally, the classification results are interpreted in accordance with actual application scenarios.

As shown in the figure above, the wheat-corn rotation area represents a hotspot of oscillation in the NDVI vegetation index during the 2022-2023 quarter, indicating significant crop growth throughout the time period, but with intervals. This is interpreted as the wheat emergence period and the corn emergence period after wheat harvest, thus classifying it as a two-season planting area. The single-season crop area indicates that the region does not exhibit significant spatiotemporal variation characteristics, and it is determined that most of the area is single-season planting. Partial fallow areas represent areas where no crops are planted for most of the planting season, with possible single-season crop planting at certain times. Fallow areas represent areas where no crops are planted throughout the entire planting season.

By using time-series remote sensing images, it is possible to determine crop phenological stages and assess agricultural activities on plots based on these stages, thereby achieving more detailed classification and enabling growth monitoring of plots based on agricultural activities and planting processes.