Movement state estimation device, movement state estimation method and program recording medium

Abstract

[Problem] To provide a motion condition estimation device, a motion condition estimation method and a motion condition estimation program capable of accurately estimating the motion condition of monitored subjects even in a crowded environment. [Solution] A motion condition estimation device according to the present invention is provided with a quantity estimating means and a motion condition estimating means. The quantity estimating means uses a plurality of chronologically consecutive images to estimate a quantity of monitored subjects for each local region in each image. The motion condition estimating means estimates the motion condition of the monitored subjects from chronological changes in the quantities estimated in each local region.

1 . A movement state estimation device comprising:
 at least one memory storing instructions; and   at least one processor executing the instructions to perform:
 estimating a number of people in a plurality of regions of an image by using machine learning; and 
 estimating a direction of flow for the image based on the estimated number of the people in the plurality of regions. 
   
     
     
         2 . The movement state estimation device according to  claim 1 , wherein the at least one processor further executes the instructions to perform:
 estimating the direction of flow for the image based on a movement of the people in the plurality of regions.   
     
     
         3 . The movement state estimation device according to  claim 2 ,
 wherein the movement of the people is based on a motion vector of pixels among a plurality of images including the image.   
     
     
         4 . The movement state estimation device according to  claim 1 ,
 wherein using the machine learning to estimate the number of the people comprises using a classifier which is trained based on crowd patches, and   wherein the crowd-patches are local images including a crowd state of an overlap of the people.   
     
     
         5 . A movement state estimation method comprising:
 estimating a number of people in a plurality of regions of an image by using machine learning; and   estimating a direction of flow for the image based on the estimated number of the people in the plurality of regions.   
     
     
         6 . The movement state estimation method according to  claim 5 , further comprising:
 estimating the direction of flow for the image is based on a movement of the people in the plurality of regions.   
     
     
         7 . The movement state estimation method according to  claim 6 ,
 wherein the movement of the people is based on a motion vector of pixels among a plurality of images including the image.   
     
     
         8 . The movement state estimation method according to  claim 5 ,
 wherein using the machine learning to estimate the number of the people comprises using a classifier which is trained based on crowd patches, and   wherein the crowd-patches are local images including a crowd state of an overlap of the people.   
     
     
         9 . A non-transitory program recording medium recording a program for causing a computer to execute:
 estimating a number of people in a plurality of regions of an image by using machine learning; and   estimating a direction of flow for the image based on the estimated number of the people in the plurality of regions.   
     
     
         10 . The non-transitory program recording medium according to  claim 9 , wherein the program causes the computer to further execute:
 estimating the direction of flow for the image is based on a movement of the people in the plurality of regions.   
     
     
         11 . The non-transitory program recording medium according to  claim 10 ,
 wherein the movement of the people is based on a motion vector of pixels among a plurality of images including the image.   
     
     
         12 . The non-transitory program recording medium according to  claim 9 ,
 wherein the program causes the computer to further execute using the machine learning to estimate the number of the people by using a classifier which is trained based on crowd patches, and   wherein the crowd-patches are local images including a crowd state of an overlap of the people.