Machine-learning based continuous camera image triggering for quality assurance inspection processes

1. A computer-implemented method for providing quality assurance comprising:

receiving data comprising a feed of images of a plurality of objects passing in front of an inspection camera module forming part of a quality assurance inspection system;

determining a centroid of an object in each image of the feed of images;

estimating a trajectory of the centroids of the objects in each image relative to one or more preceding images;

filtering out images for an object when the estimated trajectory goes outside a corresponding image;

detecting, within each remaining image, whether an object is present within the image;

identifying, for each object, images in which the object is detected that meet predefined object representation parameters;

filtering out images in which the object is detected that do not meet the predefined object representation parameters; and

providing, after the filtering, the remaining identified images to a consuming application or process for quality assurance analysis.

2. The method of claim 1 , wherein at least a portion of the objects having at least one dimension greater than a field of view of the inspection camera module.

3. The method of claim 1 , wherein the predefined object representation parameters comprise one or more of: centeredness, trajectory, minimum object size, maximum object size, sharpness, similarity to preceding images in the feed of images, or color space.

4. The method of claim 1 , wherein the consuming application or process: loads the identified images into memory, stores the identified images in physical persistence, causes the identified images to be visually displayed in a graphical user interface, or transmits the identified images to a remote computing system.

5. The method of claim 1 , wherein the detecting comprises: generating, using at least one machine learning model, a mask that indicates, via a probability on a per pixel basis, a likely location of an object within each image.

6. The method of claim 5 , wherein the at least one machine learning model is a convolutional neural network that is trained using a dataset with per pixel labels indicating locations of salient objects within images.

7. The method of claim 6 further comprising:

generating, based on the output of the convolutional neural network, a binarized image to indicate pixels within each image at which an object was detected.

8. The method of claim 7 , wherein the generating of the binarized image further comprises:

changing pixels within a predefined Hue Saturation Value (HSV) range to a first value; and

changing pixels outside the predefined HSV range to a second value.

9. The method of claim 1 further comprising:

determining an image in a trajectory of the object in which a centroid of an object is closest to a center of the image.

10. The method of claim 1 , wherein the detecting is constrained within a region of interest of each image.

11. The method of claim 1 further comprising: filtering out images which are similar to or unchanged from an image of a same object.

12. A system for providing quality assurance comprising:

at least one data processor; and

memory storing instructions which, when executed by the at least one data processor, result in operations comprising:

receiving data comprising a feed of images of a plurality of objects passing in front of an inspection camera module forming part of a quality assurance inspection system;

determining a centroid of an object in each image;

estimating a trajectory of the centroids of the objects in each image relative to one or more preceding images;

filtering out images for an object when the estimated trajectory goes outside a corresponding image;

detecting, within each remaining image, whether an object is present within the image;

identifying, for each object, all images in which the object is detected that meet predefined object representation parameters; and

filtering out images in which the object is detected that do not meet the predefined object representation parameters; and

providing, after the filtering, the remaining identified images to a consuming application or process for quality assurance analysis.

13. A computer-implemented method for providing quality assurance comprising:

receiving data comprising a feed of images of a plurality of objects passing in front of an inspection camera module forming part of a quality assurance inspection system;

determining a centroid of an object in each image in the feed of images;

estimating a trajectory of the centroids of the objects in each image relative to one or more preceding images;

filtering out images for an object when the estimated trajectory goes outside a corresponding image;

detecting, on an object-by-object basis for each remaining image, a presence of the objects having a size within a predefined range; and

filtering out images in which the object is detected that do not have a size within the predefined range; and

providing, on an object-by-object basis and after the filtering, each remaining image in which the presence of the object is detected to a consuming application or process for quality assurance analysis.

14. The method of claim 13 , wherein at least a portion of the objects having at least one dimension greater than a field of view of the inspection camera module.

15. The method of claim 13 , wherein the consuming application or process: loads the identified images into memory, stores the identified images in physical persistence, causes the identified images to be visually displayed in a graphical user interface, or transmits the identified images to a remote computing system.

16. The method of claim 13 , wherein the detecting comprises: generating, using at least one machine learning model, a mask that indicates, via a probability on a per pixel basis, a likely location of an object within each image.

17. The method of claim 16 , wherein the at least one machine learning model is a convolutional neural network that is trained using a dataset with per pixel labels indicating locations of salient objects within images.

18. The method of claim 17 further comprising:

generating, based on the output of the convolutional neural network, a binarized image to indicate pixels within each image at which an object was detected.

19. The method of claim 18 , wherein the generating of the binarized image further comprises:

changing pixels within a predefined Hue Saturation Value (HSV) range to a first value; and

changing pixels outside the predefined HSV range to a second value.