Advising image acquisition based on existing training sets

Patent metadata

Field	Value
Publication number	US-10958828-B2
Application number	US-201816157057-A
Country	US
Kind code	B2
Filing date	Oct 10, 2018
Priority date	Oct 10, 2018
Publication date	Mar 23, 2021
Grant date	Mar 23, 2021

Abstract

A method for improving the performance of a computer vision system includes obtaining input specifying a task to be performed by the computer vision system; obtaining a first digital image; and comparing the first digital image to at least one training image used to train the computer vision system to solve the task. Further steps include, based on the comparing indicating that the first digital image is insufficiently similar to the at least one training image, recommending at least one adjustment to the digital image; obtaining a second digital image in accordance with the adjustment; and performing the task with the computer vision system based on the second digital image obtained in accordance with the adjustment. Adjustments can be based, for example, on image composition and/or weather conditions.

First claim

What is claimed is:

1. A method for improving the performance of a computer vision system, said method comprising:
obtaining input specifying a task to be performed by said computer vision system;
training said computer vision system to solve said task using a plurality of training images;
determining whether current weather conditions are appropriate, based on weather conditions when at least one of said training images was taken;
responsive to said determining indicating that said current weather conditions are not appropriate, recommending a subsequent time for obtaining a first digital image;
obtaining said first digital image;
comparing said first digital image to at least one of said training images used to train said computer vision system to solve said task;
based on said comparing indicating that said first digital image is insufficiently similar to said at least one of said training images, recommending at least one adjustment to a configuration for photographically acquiring said digital image;
photographically acquiring a second digital image in accordance with said adjustment; and
performing said task with said computer vision system based on said second digital image obtained in accordance with said adjustment, wherein the recommended adjustment minimizes a difference between a histogram of the first digital image and at least one of the training images up to a first minimal threshold value and wherein the recommended adjustment minimizes a difference between a probability that the first digital image contains occluded objects and a same probability computed for the training images up to a second minimal threshold value.

2. The method of claim 1 , wherein said first digital image is obtained at said subsequent time.

3. The method of claim 1 , wherein said current weather conditions are measured by at least one of cloud cover, moisture, temperature, luminosity, and time of day.

4. The method of claim 1 , wherein said first digital image is obtained in response to said determining indicating that said current weather conditions are appropriate.

5. The system of claim 1 , further comprising associating said current weather conditions with said first digital image.

6. The method of claim 1 , wherein said current weather conditions are measured by at least one of cloud cover, moisture, temperature, luminosity, and time of day.

7. The method of claim 1 , wherein said input specifies at least one of crop identification, fruit counting, agricultural disease prediction, fruit size estimation, fruit quality estimation, fruit ripening estimation, crop pest identification, agricultural disease diagnosis, and leaf size estimation.

8. The method of claim 1 , wherein said adjustment comprises at least one of pan, tilt, and zoom.

9. The method of claim 1 , wherein recommending said at least one adjustment to said digital image comprises displaying directions to a human user operating at least one of a camera and an unmanned aerial vehicle.

10. The method of claim 1 , wherein a camera is installed on an unmanned aerial vehicle and wherein said recommendation comprises a second recommendation to adjust at least one of pitch, yaw, roll, and translation for said unmanned aerial vehicle and wherein recommending said at least one adjustment to said digital image comprises wirelessly controlling said unmanned aerial vehicle.

11. A non-transitory computer readable medium comprising computer executable instructions which when executed by a computer cause the computer to perform a method for improving the performance of a computer vision system, said method comprising:
obtaining input specifying a task to be performed by said computer vision system;
training said computer vision system to solve said task using a plurality of training images;
determining whether current weather conditions are appropriate, based on weather conditions when at least one of said training images was taken;
responsive to said determining indicating that said current weather conditions are not appropriate, recommending a subsequent time for obtaining a first digital image;
obtaining said first digital image;
comparing said first digital image to at least one of said training images used to train said computer vision system to solve said task;
based on said comparing indicating that said first digital image is insufficiently similar to said at least one of said training images, recommending at least one adjustment to a configuration for photographically acquiring said digital image;
photographically acquiring a second digital image in accordance with said adjustment; and
performing said task with said computer vision system based on said second digital image obtained in accordance with said adjustment, wherein the recommended adjustment minimizes a difference between a histogram of the first digital image and at least one of the training images up to a first minimal threshold value and wherein the recommended adjustment minimizes a difference between a probability that the first digital image contains occluded objects and a same probability computed for the training images up to a second minimal threshold value.

12. The non-transitory computer readable medium of claim 11 , wherein said first digital image is obtained at said subsequent time.

13. A computer vision system comprising:
a memory;
an image receiver;
at least one processor, coupled to said memory and said image receiver, and operative to:
obtain input specifying a task to be performed by said computer vision system;
train said computer vision system to solve said task using a plurality of training images;
determine whether current weather conditions are appropriate, based on weather conditions when at least one of said training images was taken;
responsive to said determining indicating that said current weather conditions are not appropriate, recommend a subsequent time for obtaining a first digital image;
obtain said first digital image via said image receiver;
compare said first digital image to at least one of said training images used to train said computer vision system to solve said task;
based on said comparing indicating that said first digital image is insufficiently similar to said at least one of said training images, recommend at least one adjustment to a configuration for photographically acquiring said digital image;
photographically acquire a second digital image, via said image receiver, in accordance with said adjustment; and
perform said task with said computer vision system based on said second digital image obtained in accordance with said adjustment, wherein the recommended adjustment minimizes a difference between a histogram of the first digital image and at least one of the training images up to a first minimal threshold value and wherein the recommended adjustment minimizes a difference between a probability that the first digital image contains occluded objects and a same probability computed for the training images up to a second minimal threshold value.

14. The computer vision system of claim 13 ,
wherein said first digital image is obtained at said subsequent time.

15. The computer vision system of claim 14 , wherein said current weather conditions are measured by at least one of cloud cover, moisture, temperature, luminosity, and time of day.

16. The computer vision system of claim 13 ,
wherein said first digital image is obtained in response to said determining indicating that said current weather conditions are appropriate.

17. The computer vision system of claim 16 , wherein said at

Assignees

• IBM

Inventors

• Stelmar Netto Marco Aurelio
• Britto Mattos Lima Andrea
• Malfiza Garcia De Macedo Maysa
• Zortea Maciel
• Oliveira Igor Cerqueira

Classifications

• H04N23/69

  Control of means for changing angle of the field of view, e.g. optical zoom objectives or electronic zooming · CPC title

• H04N23/695

  Control of camera direction for changing a field of view, e.g. pan, tilt or based on tracking of objects · CPC title

• B64U2101/30

  for imaging, photography or videography · CPC title

• H04N23/64 Primary

  Computer-aided capture of images, e.g. transfer from script file into camera, check of taken image quality, advice or proposal for image composition or decision on when to take image · CPC title

• G06F18/22

  Matching criteria, e.g. proximity measures · CPC title