Systems and methods for visual inspection of signal lights at railroad crossings
A system for visual inspection of signal lights at a railroad crossing includes a data source comprising a stream of images, the stream of images including images of signal lights at a railroad crossing, an inspection module configured via computer executable instructions to receive the stream of images, detect signal lights and light instances in the stream of images, encode global information relevant to ambient luminosity and return a first feature vector, encode patch information of luminosity of detected signal lights and return a second feature vector, concatenate the first feature vector with the second feature vector and return a concatenated feature vector, and decode the concatenated feature vector and provide status of the signal lights.
1. A system for visual inspection of signal lights at a railroad crossing, the system comprising:
a data source comprising a stream of images, the stream of images including images of signal lights at a railroad crossing,
an inspection module configured via computer executable instructions to receive the stream of images,
detect signal lights and light instances in the stream of images,
encode global information relevant to ambient luminosity and return a first feature vector,
encode patch information of luminosity of detected signal lights and return a second feature vector,
concatenate the first feature vector with the second feature vector and return a concatenated feature vector, and
decode the concatenated feature vector and provide status of the signal lights.
2. The system of claim 1 , further comprising:
at least one video recording device, wherein the stream of images is recorded by the at least one video recording device.
3. The system of claim 2 ,
wherein the at least one video recording device comprises one or more video cameras, configured to be installed at the railroad crossing.
4. The system of claim 1 ,
wherein the inspection module utilizes a first convolution neural network (CNN) to encode the global information relevant to ambient luminosity and return a first feature vector.
5. The system of claim 4 ,
wherein the first CNN is trained with full video frames to extract features relevant to the ambient luminosity.
6. The system of claim 1 ,
wherein the inspection module utilizes a second CNN to encode patch information of luminosity of detected signal lights and light instances and return a second feature vector.
7. The system of claim 6 ,
wherein the second CNN is trained with image patches to extract features of detected light instances.
8. The system of claim 1 ,
wherein the inspection module utilizes one or more CNN(s), wherein the CNN(s) are trained with a triplet loss function.
9. The system of claim 1 ,
wherein the inspection module utilizes a decoder network to decode the concatenated feature vector and provide status of the signal lights.
10. The system of claim 9 ,
wherein the decoder network is a third CNN.
11. A method for visual inspection of signal lights at a railroad crossing, the method comprising, through operation of at least one processor and at least one memory:
receiving a stream of images,
detecting signal lights and light instances in the stream of images,
encoding global information relevant to ambient luminosity and return a first feature vector,
encoding patch information of luminosity of detected signal lights and return a second feature vector,
concatenating the first feature vector with the second feature vector and return a concatenated feature vector, and
decoding the concatenated feature vector and providing status of the signal lights.
12. The method of claim 11 , further comprising:
recording the stream of images by at least one video recording device.
13. The method of claim 12 ,
wherein the at least one video recording device comprises one or more video cameras, configured to be installed at the railroad crossing.
14. The method of claim 11 ,
wherein the encoding of the global information comprises utilizing a first convolution neural network (CNN).
15. The method of claim 14 , further comprising:
training the first CNN with full video frames to extract features relevant to the ambient luminosity.
16. The method of claim 15 ,
wherein the training comprises utilizing a triplet loss function.
17. The method of claim 11 ,
wherein the encoding of the patch information comprising utilizing a second CNN.
18. The method of claim 17 , further comprising:
training the second CNN with image patches to extract features of detected light instances.
19. The method of claim 18 ,
wherein the training comprises utilizing a triplet loss function.
20. A non-transitory computer readable medium storing executable instructions that when executed by a computer perform a method for visual inspection of signal lights at a railroad crossing as claimed in claim 11 .