The purpose of Object Detection is to automatically determine and identify the categories and locations of multiple objects in remote sensing images based on neural network models, and mark the objects in the form of vector rectangles for subsequent spatial statistics and analysis. Compared with traditional methods, Object Detection features faster processing speed and higher accuracy. It is commonly used to rapidly determine the quantity and distribution of various object types within images.
Main Parameters
The following figure illustrates the Object Detection result of greenhouses in a specific area. The yellow bounding boxes indicate detected greenhouses, which can efficiently and accurately identify their spatial distribution, locations, and quantities, providing foundational data support for greenhouse management and cultivated land regulation.
- Function Entry: Toolbox-> Machine Learning-> Imagery Analysis-> Object Detection tool.
- File Type: Supports selecting Dataset or Folder. Batch processing of data improves inference efficiency. Default value is Dataset.
- Datasource/Dataset: Required when File Type selects Dataset. Choose the target image or Mosaic Dataset for detection.
- File Path: Required when File Type selects Folder. Specify the directory path containing image files (*.tif, *.img, etc.) to be processed.
- Model File: Select an Object Detection model file (*.sdm).
- Probability Threshold: The system calculates confidence scores for detected objects. Only results exceeding this threshold will be retained. Default is 0.5.
- Deduplicate Threshold: Typically ranges from 0.3 to 0.7 (default: 0.3). The Non-Maximum Suppression (NMS) algorithm filters overlapping candidate boxes based on Intersection over Union (IoU), removing those exceeding this threshold.
- Single Step Operation Amount: Number of images processed per inference batch. Default is 1. Higher values increase memory/VRAM usage but reduce processing time within reasonable limits.
- Processor Type: Choose between Central Processing Unit (CPU) or Graphics Processing Unit (GPU). GPU enables faster computation.
- GPU Number: Specify the GPU device ID for processing. Default is 0. Supports multi-GPU inference.
- Other Parameter Settings: Enable to configure Bounds Dataset/Datasource parameters.
- Bounds Dataset/Data source: Enables Custom Bounds inference using specified Vector Dataset extents.
- Result Data: Specifies output Region Dataset parameters. Set target Datasource and dataset name for saving detection results.
- Run: Click the Run button to execute Object Detection. Progress information will display in the Python console.
Related Content
Machine Learning Environment Configuration