Estimating a number of containers by digital image analysis

Abstract

A computer-implemented method, a computer system, and a computer program product are provided for estimating output data that includes or is based on a number of containers, which may or may not be in receptacles. The method involves using an object detection algorithm operating on an input digital image to detect container images and receptacle images (if any), and using an estimation algorithm operating on the detected images to estimate the number of containers. Estimating the number of containers may involve counting the number of containers within different container classes, as determined by the object detection algorithm. Estimating the number of containers may involve estimating a size of a receptacle in a detected receptacle image based on analysis of a detected reference container image. A detected reference object image may be used to assist with classifying detected container images in different container classes.

1 . A computer-implemented method for estimating output data comprising or based on a number of containers within a plurality of container classes, the method comprising the steps of:
 (a) using an object detection algorithm operating on an input digital image to detect at least one container image and to classify each of the at least one detected container image in one of the plurality of container classes, wherein the object detection algorithm is trained using sample images of containers each of which is annotated with one of the container classes; and   (b) using an estimation algorithm to count the number of detected container images classified in each of the container classes to estimate the output data.   
     
     
         2 . The method of  claim 1 , wherein the object detection algorithm is trained using an artificial neural network operating on the sample images of containers each of which is annotated with one of the container classes. 
     
     
         3 . The method of  claim 1 , comprising the further steps of:
 (a) receiving a confirmation that the output data is satisfactory; and   (b) further training the object detection algorithm using the input digital image and the confirmation.   
     
     
         4 . The method of  claim 1 , wherein classifying the at least one detected container image in one of the plurality of container classes is based on a pixel dimension of the container in the container image. 
     
     
         5 . The method of  claim 4 , wherein the input digital image is captured while the containers are on a moving conveyor belt having a reference marking, and wherein classifying the at least one detected container image in one of the plurality of container classes is based on a comparison of a pixel dimension of the reference marking in the input digital image, and the pixel dimension of the container in the container image. 
     
     
         6 . The method of  claim 1 , comprising the further step of displaying the output data on a display device. 
     
     
         7 . The method of  claim 1 , wherein the output data based on the number of containers comprises a monetary value of the containers, and the method comprises the further step of calculating the monetary value of the containers based on the counted number of container images classified in each of the container classes, and a stored monetary value per container for each of the container classes. 
     
     
         8 . A system for estimating output data comprising or based on a number of containers within a plurality of container classes, the system comprising:
 a computer processor; and   a computer memory comprising a non-transitory computer readable medium storing a set of instructions executable by the computer processor to implement a method comprising the steps of:   (a) using an object detection algorithm operating on an input digital image to detect at least one container image and to classify each of the at least one detected container image in one of the plurality of container classes, wherein the object detection algorithm is trained using sample images of containers each of which is annotated with one of the container classes; and   (b) using an estimation algorithm to count the number of detected container images classified in each of the container classes to estimate the output data.   
     
     
         9 . The system of  claim 8 , wherein the object detection algorithm is trained using an artificial neural network operating on the sample images each of which is annotated with one of the container classes. 
     
     
         10 . The system of  claim 8 , wherein the method comprises the further steps of:
 (a) receiving a confirmation that the output data is satisfactory; and   (b) further training the object detection algorithm using the input digital image and the confirmation.   
     
     
         11 . The system of  claim 8 , wherein classifying the at least one detected container image in one of the plurality of container classes is based on a pixel dimension of the container in the container image 
     
     
         12 . The system of  claim 11 , further comprising:
 (a) a conveyor belt having a reference marking; and   (b) a digital camera for capturing the input digital image while the containers are on a moving conveyor belt having the reference marking; and   wherein classifying the at least one detected container image in one of the plurality of container classes is based on a comparison of a pixel dimension of the reference marking in the input digital image, and the pixel dimension of the container in the container image.   
     
     
         13 . The system of  claim 8 , wherein the method comprises the further step of displaying to the output data on a display device. 
     
     
         14 . The system of  claim 8 , wherein the output data based on the number of containers comprises a monetary value of the containers, and the method comprises the further step of calculating the monetary value of the containers based on the counted number of container images classified in each of the container classes, and a stored monetary value per container for each of the container classes. 
     
     
         15 . A computer program product for estimating output data comprising or based on a number of containers, the computer program product comprising a non-transitory computer readable medium storing a set of instructions executable by a processor to implement a method comprising the steps of:
 (a) using an object detection algorithm operating on an input digital image to detect at least one container image and to classify each of the at least one detected container image in one of the plurality of container classes, wherein the object detection algorithm is trained using sample images each of which is annotated with one of the container classes; and   (b) using an estimation algorithm operating on the at least one detected container image to count the number of detected container images classified in each of the container classes.   
     
     
         16 . The computer program product of  claim 15 , wherein the object detection algorithm is trained using an artificial neural network operating on the sample images each of which is annotated with one of the container classes. 
     
     
         17 . The computer program product of  claim 15 , wherein the method comprises the further steps of:
 (a) receiving a confirmation that the output data is satisfactory; and   (b) further training the object detection algorithm using the input digital image and the confirmation.   
     
     
         18 . The computer program product of  claim 15 , wherein classifying the at least one detected container image in one of the plurality of container classes is based on a pixel dimension of the container in the container image 
     
     
         19 . The computer program product of  claim 18 , wherein the input digital image is captured while the containers are on a moving conveyor belt having a reference marking, and wherein classifying the at least one detected container image in one of the plurality of container classes is based on a comparison of a pixel dimension of the reference marking in the input digital image, and the pixel dimension of the container in the container image. 
     
     
         20 . The computer program product of  claim 15 , wherein the method comprises the further step of displaying to the output data on a display device. 
     
     
         21 . The computer program product of  claim 15 , wherein the output data based on the number of containers comprises a monetary value of the containers, and the method comprises the further step of calculating the monetary value of the containers based on the counted number of container images classified in each of the container classes, and a stored monetary value per container for each of the container classes.