Axis Deep Learning Radar Tested (D2110-VE)

By Ilya Ignatenko and Rob Kilpatrick, Published Jun 17, 2021, 10:11am EDT

Axis has released their second generation network radar, the D2110-VE, with deep-learning classification of vehicles and people. But how does it work in real world scenes?

IPVM Image

We bought and tested the D2110-VE and in this report, we examine:

  • Are vehicles and people accurately classified?
  • Does radar false alert on animals, rain, or foliage?
  • How do reflective objects impact detection?
  • Does camera tracking accurately follow objects detected by radar?
  • How does radar integrate into VMSes?
  • How does performance compare to video analytics?

Executive *******

***** ** **** * week ** ******* ** 2 *********, **** *****-** radar's *********** *** *****. On *** ******** ****, radar *** *** ***/~***' human *** ***/~***' ******* detection ****** ****** * full ***° ***** ** view, ********* ****** ***** rain, *** *** *********** was ****** ** *** radar ******* ** ** IP ****** ******* ** requiring ******** *******.

*******, *********** **** ****** structures, **** ** ******** buildings, ** ***** ******** caused ********** *** ******* which ********* ***** ****** in *** *****, ** did ***** *** ***** large ********** *******/******* ** the ***** ** ****. Both ** ***** ****** could ** ******* ***** on **** *************** (********* zones *** *********** ***********), but *** ***** **********.

*******, ****** ************** *** mostly ********, ****** **** identified ** ******** ******** times *** *** ** testing, ***** *** **** to ****** ********** ** alerts ** **** *** object **** *** *******.

Versus *****-********** **** *****

** ****-***** *******, *** D2110-VE *** ******* ** accuracy ** ****' ***** radar (*****-**), **** **** key *********:

  • **** ****** **************:*** *****-** ********** ******* as ***** ** *******, while *** *****-** *** not. ***** *** ******** unit ***** ** *** to ***** ** *****-***** or *******-***** *******, ** was ***** ** * size ******** **** *** did *** *** ********* for **************.
  • ****** ******** ****:*** *** *****-** ****** a ***** ** ~***' across * ***° ********* angle, ***** *** ******** D2050-VE ******* ***° ** ~164'.

** *** **** **** of *** *****-**, ** *** *** test *********** *** ** adjacent ********** ** ** did ** **** ****, only **** ***** ***********, so ** ****** ******* the ***. ** *** note **** ****** ******** in *** ***** ** view ****** ******* *** reflections.

Versus ***** *********

******* ****-********** ** ********* (*** *** ********) ***** ***** ***** rejection ******* ** *** Axis *****, ****** ***** non-AI ******** ******* *** more *********** ** ***** alerts, ********* ****, *******, shadows, *** ******, ***** radar **** *** "***."

*** **** ********* ****' D2110-VE *** **** ***** analytics ** ********. ***** many ******* *** ******* of ********* ******** ** similar *********, ******* ********* cover **-**° ** ***** ranges, ***** *** **** radar *** * ***° field ** ****.

*******, ***** *** *****'* price (~$*,*** ***) *** the **** ** ** Axis *** ****** ** take ********* ** ****-******** features (~$*,***-*,***), ******** ******* are ****** ** ** more ****-********* ** **** applications (*********

Detection *****

** *** *****, *** radar ******** ***** ******** reliably ****** * **** 180° ***** ** **** out ** ~***/~***', *** specified *****, ***** **' further **** *** ******** D2050-VE *****. ******** **** detected ** ~***/~***', **** as *********.

IPVM Image

Decreased *****/************ ******** ** **********

**** *********** ** **** where **** ** *** shape ** * ****** is ******, **** ** the ***** ***** **** tall *****, ******** ******* less **********, ******** *** picking ** *** ******, with ***** ********* ** ~150'.

Reflection/Scatter ******

** *** *******, ** experienced *********** ****** *** to ***** **********/********** *** of *** ******** ******, as **** ** ***** box ****** ****** ** the ***** ** ****. These *********** ****** ***** objects ** ** **********, frequently ******* ***** ******. In **** *****, ****** reflections *** *** ******* alerts *** **** ***** shown ** ******* ******* in *** *****'* **, cluttering *** ****.

IPVM Image

Axis ********/***************

** **** *****, **** recommended ********* *** ***** occupied ** ********* ** other ********** **** *** radar ****, ***** ******* some ***** ****** *** to ***********, *** ****** could *** ** ******* since **** ***** ****** throughout *** ***/***** *** did *** **** ** the **** **** **** time.

**** ************ ******** ***** reflection ****** ** ***** documentation, *** ******** **** radar ** ******** *** more **** *****:

5.5 ************ **************

**** ***** ******* *** intended *** ********** **** areas. **** *** ********* be ******-*** ***** **** as ********** ********** ** roofs, ** ******* **** where ** ******** ** expected ***** *****.

**** **** ***** *** that **** ****** **** multiple ******** ***** *** more ******:

* **** *********** *** cause ***** **********. ** scenes **** * ********* of ********** *******, **** as ******** *** *********, the ******** *********** ** the ***** ****** *** cause ***** **********.

Swaying ******** ******

** *** *****, ******* objects **** ** ******* occasionally ****** ***** ******, even **** "****** ******* Objects" *** ******* *** sensitivity *** *********. ***** alerts ***** **** ** eliminated ** ******** ***** from *** ***** ****** zone.

****' *************** ** ** were ****** ** ****** these ***** **** *** zone ** ******* ******* were ** *****.

******* ********** ******* *** cause ***** **********. **** though *** ***** ****** can ******** ****** *** trees, ******, *** ***** that **** **** *** wind, *** ********* ********* may ** ************ ** very ***** ******* ** sudden ***** ** ****. If **** ** * problem, ** *** ** recommended ** ******* ****** zones *******.

******* ******* **** **** when *** ***** ****** swayed, *.*., **** ******* on * **** **** or ******, ***** *****. Stationary ******* ******** ** be ****** ** *** radar.

**** ************** ** **** case *** ****** ** mount *** ***** ** a **** ****** *******.

No ****** ** *******

*** *****-** ******* ******* in *** ***** ** view, **** ***** *** small, **** ** ***** alerts. ******* **** ****** displayed ** *******, *** did *** ******* ******/******* alerts, ** ****** *** detected ** ***.

*** *******, ** **** image, *** ****** ** accurately ******** (***** ** red ** *** **** while *** *** ** shown ** ******* *** not ** *****.

IPVM Image

Subjects ** ***** ********** ** *******

****** *******, ******** ****** bikes **** ********** ********** as "*******", ********** ** settings.

IPVM Image

People ********** ** ********

************, ****** **** ********* classified ** ******** ***** walking ******* *** *****. This *** *** ****** often, day *** ** ******** of ******.

IPVM Image

PTZ ****-******** ***********

*** **** *****'* ****-******** app ** *** *****-**, instead ** ** ***-** ACAP ** ******* *********** in *** **** *****. Performance *** ***** ** our *****. **** ******* were ******* **********, **** as *** *** ******* through *** ******* *** here:

IPVM Image

*******, ** ***** ***** that ** ****** ****** the ****** ****** ******* the *****, ****** ******** up (***** *****).

IPVM Image

**** ***** ** ******* by ******* ******* **** level ** * ******* focal ******, *** **** made ******* *******, ******** visible *******.

Accurate ********/** ***** ****** ** ****

*** **** ***** *** unaffected ** ****, **** detection ********** *********** *** same ** ** ***** conditions *** ** ***** alerts ****** ** **** drops (* ****** ***** in ***** *********).

IPVM Image

VMS ***********

*** *****-** **** ** VMSes ** * ******, instead ** ***** ***** integrations ** *******, *** uses * *** ****** channel *******. **** ******* the ****** ** *********** work ***** **** ** performed, ** ** **** the **** ***** ***.

IPVM Image

*******, **** ***** ***** with ******** ********* **** view **** *** **********, instead ** ***** **** to ***** **** *** on * ****** *** view ******* ***** ************. Other ********* ********* ******* (e.g.,*****) ********* **** ****** and ******** ********, *** a ***** ******, *** require ********** *** *********** to ***** ***** ******* on * *** ****, though **** **** ********* view ******** ****** ** a ****** ***.

Genetec *** ****** ***********

** ******** ** *** standard ***********, *** *****-** is ********** ** ******* via *** ********** ******** Area (***) ******, ***** allows ***** ** ****** arming ***** ** * map. **** *********, *** radar ** ***** ** a ******** ******, ***** displays ****** ******** *** track ** **** ****.

IPVM Image

**** *********** ** ******* to********'* ****** ***** ***********, *** *** **** drive *** *******, ****** alarms, ** ******* ***** events.

Physical ********/********** ** ***** *** *****

*** *****-** ** ********** very ******* ** ****' original *****-** *****, **** some *** *******:

  • ***-************** ********:*** ******** ***** **** a ***° ******** ***** could ** ************ ** the ****/***** ** ******, which ** ** ****** needed ** *** *****-** due ** *** ***° detection *****.
  • *** ***:*** *****-** ******** * PoE ****** **** *** be **** ** ***** an ** ****** ** other ******, **** ** an**** **** *******, *** ******** ** the ******** *****.

** **** ** *** device's ************ *** ***** new ******** ** **** video:

Versions ****

*** ********* ******** **** used ****** *******.

  • **** *****-**:**.*.*
  • ******* ******** ******: *.*.*.* (167.90)
  • ******* *** ******: *.*

Comments (4)

***'* *** ****** ** articles **** *****!*** ** you ***** *** **** would ******* ** * marine *********** **** ** an **** ****? *********** the **** *** *** water.

Agree
Disagree
Informative
Unhelpful
Funny

**** ********. ** ***'* have * ***** **** of ***** ** **** with ***** ** *** office, ** * ***'* want ** *********. ** might **** ** ***** testing ** **** ***** on. ** *** ******** I **** ***** **** how *** ***** ***** classify *****, ******* *'* not **** ** ****'** included ** "*******" ************** or ***.

Agree
Disagree
Informative: 4
Unhelpful
Funny

***** *******. * *** confirm *** ********** ***** is ******* ** *** D2050-VE.

Agree
Disagree
Informative: 3
Unhelpful
Funny

***** ******! *'* ** interested ** ****** *** configurational **** ** **** as ****.

* *** ***** **** see ******* ** * dock/lake; * ****** ******* from * ********* ** course. *'* ** ***** to **** ********** **** test ********* **** **** to ***** ***. ******!

Agree
Disagree
Informative
Unhelpful
Funny
Read this IPVM report for free.

This article is part of IPVM's 7,216 reports and 962 tests and is only available to subscribers. To get a one-time preview of our work, enter your work email to access the full article.

Already a subscriber? Login here | Join now
Loading Related Reports