Fusing deep learning and geometric constraint for image-based localization
A computer-implemented method, comprising applying training images of an environment divided into zones to a neural network, and performing classification to label a test image based on a closest zone of the zones; extracting a feature from retrieved training images and pose information of the test image that match the closest zone; performing bundle adjustment on the extracted feature by triangulating map points for the closest zone to generate a reprojection error, and minimizing the reprojection error to determine an optimal pose of the test image; and for the optimal pose, providing an output indicative of a location or probability of a location of the test image at the optimal pose within the environment.
1. A computer-implemented method, comprising:
applying training images of an environment divided into zones to a neural network, and performing classification to label a test image based on a closest zone of the zones;
extracting a feature from retrieved training images and pose information of the test image that match the closest zone;
performing bundle adjustment on the extracted feature by triangulating map points for the closest zone to generate a reprojection error, and minimizing the reprojection error to determine an optimal pose of the test image; and
for the optimal pose, providing an output indicative of a location or probability of a location of the test image at the optimal pose within the environment.
2. A computer-implemented method of claim 1 , wherein the applying the training images comprises receiving the training images associated with poses in zones of the environment as historical or simulation data, and providing the received training images to a neural network.
3. A computer-implemented method of claim 2 , wherein the neural network is a deep learning neural network that learns zones associated with the pose, and determines the closest zone for the test image.
4. A computer-implemented method of claim 1 , wherein the bundle adjustment comprises reprojecting the 3D points associated with a measured pose and the triangulated map points into 2D image space to generate a result, and comparing the result to a registered 2D observation to determine the reprojection error.
5. A computer-implemented method of claim 4 , wherein for reprojection error being below or equal to a threshold, the pose of the test image is confirmed to be the optimal pose.
6. A computer-implemented method of claim 4 , wherein for reprojection error being above a threshold, the pose of the test image is determined to be not correct and a calculation of the pose of the test image is determined to be correct.
7. A computer-implemented method of claim 1 , wherein the minimizing the reprojection error comprises adjusting the pose of the test image to minimize the reprojection error.
8. A non-transitory computer readable medium having a storage that stores instructions, the instructions executed by a processor, the instructions comprising:
applying training images of an environment divided into zones to a neural network, and performing classification to label a test image based on a closest zone of the zones;
extracting a feature from retrieved training images and pose information of the test image that match the closest zone;
performing bundle adjustment on the extracted feature by triangulating map points for the closest zone to generate a reprojection error, and minimizing the reprojection error to determine an optimal pose of the test image; and
for the optimal pose, providing an output indicative of a location or probability of a location of the test image at the optimal pose within the environment.
9. A non-transitory computer readable medium of claim 8 , wherein the applying the training images comprises receiving the training images associated with poses in zones of the environment as historical or simulation data, and providing the received training images to a neural network.
10. A non-transitory computer readable medium of claim 9 , wherein the neural network is a deep learning neural network that learns zones associated with the pose, and determines the closest zone for the test image.
11. A non-transitory computer readable medium of claim 8 , wherein the bundle adjustment comprises reprojecting the 3D points associated with a measured pose and the triangulated map points into 2D image space to generate a result, and comparing the result to a registered 2D observation to determine the reprojection error.
12. A non-transitory computer readable medium of claim 11 , wherein for reprojection error being below or equal to a threshold, the pose of the test image is confirmed to be the optimal pose.
13. A non-transitory computer readable medium of claim 11 , wherein for reprojection error being above a threshold, the pose of the test image is determined to be not correct and a calculation of the pose of the test image is determined to be correct.
14. A non-transitory computer readable medium of claim 8 , wherein the minimizing the reprojection error comprises adjusting the pose of the test image to minimize the reprojection error.
15. A computer-implemented system for localizing and tracking a scope in an environment to identify a target, comprising:
a memory configured to store a program; and
a processor communicatively coupled to the memory and configured to execute the program to:
apply training images of the environment that is associated with the scope, divided into zones to a neural network, and performing classification to label a test image, generated by the scope, based on a closest zone of the zones of the environment associated with the scope;
extract a feature from retrieved training images and pose information of the test image that match the closest zone;
perform bundle adjustment on the extracted feature by triangulating map points for the closest zone to generate a reprojection error, and minimizing the reprojection error to determine an optimal pose of the test image; and
for the optimal pose, provide an output indicative of a location or probability of a location of the test image generated by the scope, at the optimal pose within the environment.
16. The computer-implemented system of claim 15 , wherein the environment comprises a gastrointestinal tract, or a bronche tract of one or more lungs.
17. The computer-implemented system of claim 15 , further comprising the scope, wherein the scope is configured to provide a location of one or more targets that comprise at least one of a polyp, a lesion and a cancerous tissue.
18. The computer-implemented system of claim 15 , further comprising the scope, wherein the scope comprises one or more sensors configured to receive the test image associated with the environment, and the test image is a visual image.
19. The computer-implemented system of claim 15 , further comprising the scope, wherein the scope is an endoscope or a bronchoscope.
20. The computer-implemented system of claim 15 , wherein the environment is a piping system, a subterranean environment, or an industrial facility.