Pinno718
Legendary Member
Can't even use the quote function to see what you actually wrote...
It began with f and ended in g.
Can't even use the quote function to see what you actually wrote...
They look nice. Fair play.
Pity lot of the drones have glorious technicolour cameras (see wall poster page 302 'Highway to hell').
Someone is cashing in on black or white paint.
Anyhoo, I might be going to Ukraine in October. Fingers crossed.
Pity lot of the drones have glorious technicolour cameras (see wall poster page 302 'Highway to hell').
Someone is cashing in on black or white paint.
Anyhoo, I might be going to Ukraine in October. Fingers crossed.
They are probably trying the moiré effect due to sampling rate in the cameras. If the spatial period of the bands is chosen carefully, it will look nothing like a lorry in the image. This is from the same fundamental reason that in TV shirts with vertical stripes are discouraged, as you ger all sorts of strange patterns due to sampling. It is less of an issue with HD cameras and with high frame rate, but that may not be what drones use.
I'm taking you to Ukraine.
I'm taking you to Ukraine.
They are probably trying the moiré effect due to sampling rate in the cameras. If the spatial period of the bands is chosen carefully, it will look nothing like a lorry in the image. This is from the same fundamental reason that in TV shirts with vertical stripes are discouraged, as you ger all sorts of strange patterns due to sampling. It is less of an issue with HD cameras and with high frame rate, but that may not be what drones use.
I've not read beyond aspects raised here but those zebra vehicles look like operating close to front. Also I thought Ukraine was increasingly using fibre linked drones to avoid RF links being blocked. I would have expected fibre links to provide a higher data bandwidth allowing processing to be based with the controller rather than on the drone itself.Vision AI often processes the video for edge detection as this helps extract "features" to classify objects. This is especially the case with things like drones where the processing power available is going to be much lower than in the cloud, which would also probably require the input video to be significantly reduced in resolution. Low power AI will only support more limited feature extraction.
Interesting that they mention about interfering with the machine vision of UAVs. Ukraine has started to employ AI enabled drones for automatic targeting and had considerable success. Vision AI often processes the video for edge detection as this helps extract "features" to classify objects. This is especially the case with things like drones where the processing power available is going to be much lower than in the cloud, which would also probably require the input video to be significantly reduced in resolution. Low power AI will only support more limited feature extraction.
I am guessing these stripes introduce a whole load of edges to the video image being processed and this could confuse the AI models being used. Reports say that the AI has been trained on thousands of hours of video featuring Russian vehicles, presumably much in more standard camouflage.
If this is the case, then it is a cheap way to try to better protect the vehicles, but may only be temporary if Ukraine can modify its models and training. There must be a limit between fooling the AI and standing out in a battlefield to humans. Plus I believe there can be multiple different sensor types (eg. LIDAR, infrared etc) that may be used alongside visual video with AI to negate the camouflage.