Artificial intelligence based plantable blank spot detection

1. An artificial intelligence based plantable blank spot detection apparatus comprising:

a model identifier, executed by at least one hardware processor, to

generate a plurality of clusters of input images of areas that are to be analyzed for plantable blank spot detection;

identify, for each cluster of the plurality of clusters, a model to analyze corresponding images of a cluster;

select, from the models identified for the plurality of clusters, a model to analyze the input images; and

analyze, based on the selected model, the input images;

a generative adversarial analyzer, executed by the at least one hardware processor, to

identify canal lines in the analyzed images;

an image classifier, executed by the at least one hardware processor, to

determine, based on the canal lines identified in the analyzed images, plantable blank spots; and

a drone controller, executed by the at least one hardware processor, to

control an operation of a drone to validate the determination of the plantable blank spots.

2. The apparatus according to claim 1 , wherein the model identifier is executed by the at least one hardware processor to generate the plurality of clusters of input images of areas that are to be analyzed for plantable blank spot detection by:

generating the plurality of clusters based on an age of vegetation including trees, a species of the vegetation, and a land type of the areas that are to be analyzed for plantable blank spot detection.

3. The apparatus according to claim 1 , wherein the model identifier is executed by the at least one hardware processor to select, from the models identified for the plurality of clusters, the model to analyze the input images by:

determining, for each cluster of the plurality of clusters, from a plurality of available models, the selected model based on a highest number of matches to an age of vegetation including trees in the corresponding images of the cluster, a species of the vegetation, and a land type of the areas that are to be analyzed for plantable blank spot detection.

4. The apparatus according to claim 1 , wherein the generative adversarial analyzer is executed by the at least one hardware processor to identify canal lines in the analyzed images by:

implementing CycleGAN to identify canal lines in the analyzed images.

5. The apparatus according to claim 1 , wherein the generative adversarial analyzer is executed by the at least one hardware processor to identify canal lines in the analyzed images by:

utilizing a generator to

learn features of a real-world input image, and

translate the input image to a generated image;

utilizing a discriminator to distinguish between the real-world input image and the generated image;

minimizing a loss factor with respect to the generator and the discriminator; and

identifying, based on images generated by the generator from the analyzed images, the canal lines in the images generated by the generator.

6. The apparatus according to claim 1 , wherein the image classifier is executed by the at least one hardware processor to determine, based on the canal lines identified in the analyzed images, plantable blank spots by:

utilizing a convolutional neural network (CNN) to determine, based on the canal lines identified in the analyzed images, plantable blank spots.

7. The apparatus according to claim 1 , wherein the image classifier is executed by the at least one hardware processor to determine, based on the canal lines identified in the analyzed images, plantable blank spots by:

utilizing a convolutional neural network (CNN) to distinguish between a flood area, a tree area, a weed area, and a plantable blank spot in each image of the analyzed images.

8. The apparatus according to claim 1 , further comprising:

an image pre-processor, executed by the at least one hardware processor, to generate texture features for the analyzed images.

9. The apparatus according to claim 1 , further comprising:

an image post-processor, executed by the at least one hardware processor, to perform canal contour extraction for the analyzed images.

10. The apparatus according to claim 1 , further comprising:

an image post-processor, executed by the at least one hardware processor, to perform morphological operations of dilation, erosion, and subtraction for the analyzed images.

11. The apparatus according to claim 1 , further comprising:

an image post-processor, executed by the at least one hardware processor, to perform skeletonisation to connect individual canal contours for the analyzed images.

12. A method for artificial intelligence based plantable blank spot detection, the method comprising:

generating, by at least one hardware processor, a plurality of clusters of input images of areas that are to be analyzed for plantable blank spot detection;

identifying, by the at least one hardware processor, for each cluster of the plurality of clusters, a model to analyze corresponding images of a cluster;

selecting, by the at least one hardware processor, from the models identified for the plurality of clusters, a model to analyze the input images;

analyzing, by the at least one hardware processor, based on the selected model, the input images;

generating, by the at least one hardware processor, texture features for the analyzed images;

identifying, by the at least one hardware processor, canal lines in the analyzed images, the analyzed images including the generated texture features;

determining, by the at least one hardware processor, based on the canal lines identified in the analyzed images, plantable blank spots; and

controlling, by the at least one hardware processor, an operation of a drone to validate the determination of the plantable blank spots.

13. The method according to claim 12 , wherein generating, by the at least one hardware processor, the plurality of clusters of input images of areas that are to be analyzed for plantable blank spot detection, further comprises:

generating the plurality of clusters based on an age of vegetation including trees, a species of the vegetation, and a land type of the areas that are to be analyzed for plantable blank spot detection.

14. The method according to claim 12 , wherein selecting, by the at least one hardware processor, from the models identified for the plurality of clusters, the model to analyze the input images, further comprises:

determining, for each cluster of the plurality of clusters, from a plurality of available models, the selected model based on a highest number of matches to an age of vegetation including trees in the corresponding images of the cluster, a species of the vegetation, and a land type of the areas that are to be analyzed for plantable blank spot detection.

15. The method according to claim 12 , wherein identifying, by the at least one hardware processor, canal lines in the analyzed images, further comprises:

implementing CycleGAN to identify, canal lines in the analyzed images.

16. The method according to claim 12 , wherein identifying, by the at least one hardware processor, canal lines in the analyzed images, further comprises:

utilizing a generator to

learn features of a real-world input image, and

translate the input image to a generated image;

utilizing a discriminator to distinguish between the real-world input image and the generated image;

minimizing a loss factor with respect to the generator and the discriminator; and

identifying, based on images generated by the generator from the analyzed images, the canal lines in the images generated by the generator.

17. A non-transitory computer readable medium having stored thereon machine readable instructions, the machine readable instructions, when executed by at least one hardware processor, cause the at least one hardware processor to:

generate a plurality of clusters of input images of areas that are to be analyzed for plantable blank spot detection;

identify, for each cluster of the plurality of clusters, a model to analyze corresponding images of a cluster;

select, from the models identified for the plurality of clusters, a model to analyze the input images;

analyze, based on the selected model, the input images;

identify canal lines in the analyzed images;

determine, based on the canal lines identified in the analyzed images, plantable blank spots;

perform canal contour extraction for the analyzed images; and

control an operation of a drone to validate the determination of the plantable blank spots, the analyzed images including the canal contour extraction.

18. The non-transitory computer readable medium according to claim 17 , wherein the machine readable instructions to determine, based on the canal lines identified in the analyzed images, plantable blank spots, when executed by the at least one hardware processor, further cause the at least one hardware processor to:

utilize a convolutional neural network (CNN) to determine, based on the canal lines identified in the analyzed images, plantable blank spots.

19. The non-transitory computer readable medium according to claim 17 , wherein the machine readable instructions to determine, based on the canal lines identified in the analyzed images, plantable blank spots in the analyzed images, when executed by the at least one hardware processor, further cause the at least one hardware processor to:

utilize a convolutional neural network (CNN) to distinguish between a flood area, a tree area, a weed area, and a plantable blank spot in each image of the analyzed images.

20. The non-transitory computer readable medium according to claim 17 , wherein the machine readable instructions, when executed by the at least one hardware processor, further cause the at least one hardware processor to:

generate texture features for the analyzed images.