Automated work chart systems and methods
The systems and methods provide an action recognition and analytics tool for use in manufacturing, health care services, shipping, retailing and other similar contexts. Machine learning action recognition can be utilized to determine cycles, processes, actions, sequences, objects and or the like in one or more sensor streams. The sensor streams can include, but are not limited to, one or more video sensor frames, thermal sensor frames, infrared sensor frames, and or three-dimensional depth frames. The analytics tool can provide for automatic creation of work charts.
1. A method of creating work charts comprising:
receiving one or more given indicators or criteria including a selection of cycles within a given time period;
accessing one or more given data sets based on the one or more given indicators or criteria, wherein the one or more given data sets include one or more indicators of at least one of one or more cycles, one or more processes, one or more actions, one or more sequences, one or more objects, and one or more parameters of a manufacturing operation of a product determined by convolution neural network deep learning using a computing device executing a machine learning engine over each of a plurality of video frame sensor streams from a plurality of manufacturing stations across an assembly line, wherein the one or more given indicators of the at least one of one or more cycles, one or more processes, one or more actions, one or more sequences, one or more objects, and one or more parameters of the manufacturing operation are indexed to corresponding portions of the plurality of video frame sensor streams, wherein the determination by the convolution neural network deep learning comprises:
performing, with a frame feature extractor, a two-dimensional convolution operation on video frames of the plurality of video frame sensor streams to generate a two-dimensional array of feature vectors;
combining neighboring feature vectors in the two-dimensional array of feature vectors and determining a dynamic region of interest, with a region of interest detector, in a set of the neighboring feature vectors, wherein the region of interest detector and the frame feature extractor share layers of a convolution neural network; and
extracting a feature vector from an area within the dynamic region of interest to analyze with the convolution neural network while discarding remaining feature vectors of the two-dimensional array outside of the dynamic region of interest;
determining, by the computing device executing the machine learning engine, a representative data set including the at least one of one or more cycles, one or more processes, one or more actions, one or more sequences, one or more objects and one or more parameters, and quantitative data, judgement data and inference data statistically derived from the one or more given data sets for the selection of cycles within the given time period;
creating, by the computing device executing the machine learning engine, a work chart from the representative data set, wherein the work chart includes a plurality of work elements and dependencies between the plurality of work elements and associated time for performing the plurality of work elements and associated time between the plurality of work elements for the selection of cycles within the given time period, and wherein the plurality of work elements are indexed to the at least one of one or more cycles, one or more processes, one or more actions, one or more sequences, one or more objects, and one or more parameters of the manufacturing operation of the representative data set and to the corresponding portions of the plurality of video frame sensor streams; and
adjusting the manufacturing operation of the product to gain efficiencies of one or more of the plurality of work elements based on the dependencies between the plurality of work elements and the associated time for performing the plurality of work elements and the associated time between the plurality of work elements in the work chart.
2. The method according to claim 1 , wherein a subject of the work chart comprises an article of manufacture, a health care service, a shipping transaction or a retailing transaction.
3. The method according to claim 1 , wherein the work chart is organized based on a lean framework including value added and non-value added, wherein the non-value added includes idle, walking and necessary.
4. The method according to claim 1 , further comprising:
receiving a selection of a work element;
retrieving the one or more indicators of the at least one of the one or more cycles, one or more processes, one or more actions, one or more sequences, one or more objects, one or more parameters indexed by the selected work element; and
outputting the one or more indicators of the at least one of one or more of the one or more cycles, one or more processes, one or more actions, one or more sequences, one or more objects, one or more parameters, or the corresponding portions of the plurality of video frame sensor streams for the selected work element.
5. The method according to claim 4 , further comprising:
retrieving the corresponding portions of the plurality of video frame sensor streams indexed by the selected work element; and
outputting the corresponding portions of the plurality of video frame sensor streams indexed by the selected work element.
6. The method according to claim 1 , wherein the one or more given data sets and the corresponding portions of the plurality of video frame sensor streams are blockchained.
7. One or more non-transitory computing device-readable storage mediums storing instructions executable by one or more computing devices to perform an action recognition and analytics method of creating work charts comprising:
receiving one or more given indicators or criteria including a selection of cycles within a given time period;
accessing one or more given data sets based on the one or more given indicators or criteria, wherein the one or more given data sets include one or more indicators of at least one of one or more cycles, one or more processes, one or more actions, one or more sequences, one or more objects, and one or more parameters of a manufacturing operation of a product indexed to corresponding portions of one or more video sensor streams from a plurality of manufacturing stations across an assembly line, wherein the one or more indicators of the at least one of one or more cycles, one or more processes, one or more actions, one or more sequences, one or more objections and one or more parameters are determined by convolution neural network deep learning using a computing device executing a machine learning engine from the one or more video sensor streams, the determination by the convolution neural network deep learning comprising:
performing a two-dimensional convolution operation on the one or more video sensor streams with a frame feature extractor to generate a two-dimensional array of feature vectors;
determining a dynamic region of interest in the two-dimensional array of feature vectors with a region of interest detector, wherein the frame feature extractor and the region of interest extractor share convolution layers in a convolution neural network; and
extracting a fixed-size feature vector from an area of the one or more video sensor streams within the dynamic region of interest and analyzing the fixed-size feature vector with the convolution neural network without analyzing an area of the one or more video sensor streams outside the dynamic region of interest;
statistically analyzing, by the computing device executing the machine learning engine, the one or more given data sets based on the one or more given indicators or criteria to determine a representative data set including the at least one of one or more cycles, one or more processes, one or more actions, one or more sequences, one or more objects and one or more parameters, and quantitative data, judgement data and inference data for the selection of cycles within the given time period;
creating, by the computing device executing the machine learning engine, a work chart from the representative data set, wherein the work chart includes a plurality of work elements and dependencies between the plurality of work elements and associated time for performing the plurality of work elements and associated time between the plurality of work elements for the selection of cycles within the given time period, and wherein the plurality of work elements are indexed to the at least one of one or more cycles, one or more processes, one or more actions, one or more sequences, one or more objects, and one or more parameters of the manufacturing operation of the representative data set and to the corresponding portions of the one or more video sensor streams; and
adjusting one or more of the plurality of work elements of the manufacturing operation of the product to gain efficiencies based on the dependencies between the plurality of work elements and the associated time for performing the plurality of work elements and the associated time between the plurality of work elements in the work chart.
8. The one or more non-transitory computing device-readable storage mediums storing instructions executable by one or more computing devices to perform the action recognition and analytics method of creating work charts according to claim 7 , wherein a subject of the one or more given data sets comprises an article of manufacture, a health care service, a shipping transaction or a retailing transaction.
9. The one or more non-transitory computing device-readable storage mediums storing instructions executable by one or more computing devices to perform the action recognition and analytics method of creating work charts according to claim 7 , wherein the work chart is organized based on a lean framework including value added and non-value added.
10. The one or more non-transitory computing device-readable storage mediums storing instructions executable by one or more computing devices to perform the action recognition and analytics method of creating work charts according to claim 9 , wherein the non-value added includes idle, walking and necessary.
11. The one or more non-transitory computing device-readable storage mediums storing instructions executable by one or more computing devices to perform the action recognition and analytics method of creating work charts according to claim 7 , further comprising:
receiving a selection of a work element; and
outputting the one or more indicators of the at least one of one or more of the one or more cycles, one or more processes, one or more actions, one or more sequences, one or more objects, one or more parameters, or the corresponding portions of the one or more video sensor streams for the selected work element.
12. The one or more non-transitory computing device-readable storage mediums storing instructions executable by one or more computing devices to perform the action recognition and analytics method of creating work charts according to claim 11 , wherein the one or more indicators of the corresponding ones of the one or more cycles, the one or more processes, the one or more actions, the one or more sequences, the one or more objects, and the one or more parameters are indexed to the corresponding portions of the one or more video sensor streams by corresponding time stamps.
13. A system comprising:
one or more data storage units;
one or more interfaces;
one or more engines configured to:
receive one or more given indicators or criteria including a selection of cycles within a given time period;
access one or more given data sets stored on the one or more data storage units based on the one or more given indicators or criteria, wherein the one or more given data sets include one or more indicators of at least one of one or more cycles, one or more processes, one or more actions, one or more sequences, one or more objects, and one or more parameters of a manufacturing operation of a product, indexed to corresponding portions of one or more video sensor streams from a plurality of manufacturing stations across an assembly line, wherein the one or more indicators of the at least one of one or more cycles, one or more processes, one or more actions, one or more sequences, one or more objections and one or more parameters are determined by convolution neural network deep learning using a computing device executing a machine learning engine from the one or more video sensor streams, the determination by the convolution network deep learning comprising:
performing, by a frame feature extractor, a two-dimensional convolution operation on the one or more video sensor streams to generate a two-dimensional array of feature vectors;
determining, by a region of interest detector, a dynamic region of interest in the two-dimensional array of feature vectors, wherein the frame feature extractor and the region of interest detector share convolution layers of a convolution neural network; and
analyzing, by a region of interest pooling unit of the convolution neural network, an area of the one or more video sensor streams in an area of the dynamic region of interest and discarding feature vectors of the two-dimensional array of feature vectors outside the area of the dynamic region of interest;
determine, by the computing device executing the machine learning engine, a representative data set from the one or more given data sets based on a statistical analysis of the at least one of one or more cycles, one or more processes, one or more actions, one or more sequences, one or more objects and one or more parameters, and quantitative data, judgement data and inference data for the selection of cycles within the given time period;
create, by the computing device executing the machine learning engine, a work chart from the representative data set, wherein the work chart includes a plurality of work elements and dependencies between the plurality of work elements and associated time for performing the plurality of work elements and associated time between the plurality of work elements for the selection of cycles within the given time period, and wherein the plurality of work elements are indexed to the at least one of one or more cycles, one or more processes, one or more actions, one or more sequences, one or more objects, and one or more parameters of the manufacturing operation of the representative data set and to the corresponding portions of the one or more video sensor streams; and
adjust the manufacturing operation of the product to gain efficiencies of one or more of the plurality of work elements based on the dependencies between the plurality of work elements and the associated time for performing the plurality of work elements and the associated time between the plurality of work elements in the work chart.
14. The system of claim 13 , wherein the one or more given data sets and the corresponding portions of the one or more video sensor streams are blockchained.
15. The system of claim 13 , wherein a subject of the one or more video sensor streams comprises an article of manufacture, a health care service, a shipping transaction or a retailing transaction.
16. The system of claim, 13 , wherein the one or more indicators of the at least one of one or more cycles, one or more processes, one or more actions, one or more sequences, one or more objects and one or more parameters are indexed to the corresponding portions of the one or more video sensor streams by corresponding time stamps.
17. The system of claim 13 , wherein the work chart is organized based on a lean framework including value added and non-value added.
18. The system of claim 17 , wherein the non-value added includes idle, walking and necessary.
19. The system of claim 13 , wherein the one or more engines are further configured to:
receive a selection of a work element;
retrieve the one or more indicators of the at least one of the one or more cycles, one or more processes, one or more actions, one or more sequences, one or more objects, one or more parameters and the corresponding portions of the one or more video sensor streams indexed by the selected work element from the one or more given data sets stored on the one or more data storage units; and
output the one or more indicators of the at least one of one or more cycles, one or more processes, one or more actions, one or more sequences, one or more objects, one or more parameters, or the corresponding portions of the one or more video sensor streams for the selected work element in a graphical user interface on the one or more interfaces.