Modeling Guide for SAP Data Hub

Recognize by Distance

The Recognize by Distance operator recognizes the objects detected in the previous frame in the current frame by its distance shift.

This operator uses the Kalman filter to predict the object's size and position for comparison purposes. If an object with an old enough past has not been matched with a rectangle in the current frame, the operator can also use this predicted information to retrieve the missing rectangle (hence its size and position in the frame).

Configuration Parameters

Parameter

Type

Description

minAge

int

Minimum age required for the detected rectangle to be considered as an object (in number of frames).
maxSizeDelta

int

Maximum size threshold (in pixels) to determine if a past rectangle's size and a current rectangle's size are close enough to be matched to the same object.

maxDistance

int

Maximum distance threshold (in pixels) to determine if a past rectangle's position and a current rectangle's position are close enough to be matched to the same object.

maxFramePrediction

int

If greater than zero, the operator will affect a new rectangle to an old enough object using this object's past information. The number of consecutive predicted information of an object cannot exceed this value.

Input

Input

Type

Description

objectsIn

python.cv.objects The list containing the past frame objects' information.

rectsIn

message.python.image.rectangles

A list of rectangles that have been detected in the current frame.

Output

Output

Type

Description

objectsOut

python.cv.objects The updated list of objects.