A detection tool with Matlab UI.
Need to install Matlab first.
Run "main.m" to activate the app interface.
Play with the following steps:
- Click "Open New Image" to load the image you want to detect. Browse and choose the image in the pop-up window.
- Choose the segmentation method for shadow removal, and click "Segment" to apply it.
- 2.1 K-means The parameter of K-means is the number of the cluster. The default k equals 3. Small k's can avoid most of the noise in the shadow. Larger k provides more details.
- 2.2 Mean shift The parameter of Mean shift is the bandwidth of the kernel. The default bw equals to 0.2. It's a faster method for segmentation. The detection result is very similar to the K-means when k = 3.
- Click "Mask" to generate the binary mask for shadow removal.
- Click "Output" to get the detection result.
- (optional) Save the image.
- Click "Count" to count the number of detected clusters.
-
After each step you might need to wait for a few seconds to see the image change, which indicates the current step finished. The waiting time depends on the input file size and computing speed.
-
This app referenced the mean shift method from K. Fukunaga and L.D. Hosteler, "The Estimation of the Gradient of a Density Function, with Applications in Pattern Recognition". PDF
@article{valicharla2024morning, title={Morning Glory Flower Detection in Aerial Images Using Semi-Supervised Segmentation with Gaussian Mixture Models}, author={Valicharla, Sruthi Keerthi and Wang, Jinge and Li, Xin and Gururajan, Srikanth and Karimzadeh, Roghaiyeh and Park, Yong-Lak}, journal={AgriEngineering}, volume={6}, number={1}, pages={555--573}, year={2024}, publisher={MDPI} }