VL53L1X People Counting range status

[SBann.1]

Dear,

What are the best practices for the people counting algorithm regarding range status? I often get status 2, so I lowered the signal threshold to 400. This reduced it, but did not solve it completely. What status codes should I allow, and which ones are "real" errors?

Thanks in advance.

Kind regards,

Stijn Bannink

[John E KVAM]

Status 2 means you don't have enough signal too get a reading. And lowering the threshold might solve a particular case, but there will be another one. At some point the target is just too dull or two far away and the sensor does not get enough photons to get an answer. I'd treat status 2 as 'nothing there'. The assumption is that if something was there, you would have seen it.

[John E KVAM]

Status 2 means you don't have enough signal too get a reading. And lowering the threshold might solve a particular case, but there will be another one. At some point the target is just too dull or two far away and the sensor does not get enough photons to get an answer. I'd treat status 2 as 'nothing there'. The assumption is that if something was there, you would have seen it.

[Akuma.6]

Hi,

I made our custom board based on your ST VL53L1X sensor. I found some disadvantages to counting the people in and out is not properly recognising using a single VL53L1X sensor.

Is there any specific timing or any default parameters it should work? Or are there any dual sensors connected in your EVK setup.

I followed the below video link to convert the same logic in our platform. 

https://youtu.be/c91Ve-g0J2U

Please suggest to me how to get the ppl counting number?

Regards,

Kumar

[Akuma.6]

Hi,

I defined the below settings in my setup. partially working but not acceptable.

// Configure distance mode to LONG distance mode

DISTANCE_MODE = 2

// Set timing budget

timing_budget = 33

// Set region of interest center

BACK_ZONE_CENTER = 231

// Set region of interest SPADs 8x16

ROI_SPADS = 8x16

//DISTANCE MIN

min_distance = 0;

//DISTANCE MAX

max_distance = 2500;

//THRESHOLD DISTANCE

distance_threshold = 1200;

#define DISTANCES_ARRAY_SIZE   10  // number of samples

#define FRONT_ZONE_CENTER     175

#define BACK_ZONE_CENTER     231

#define PATTERN_ZONE_LEFT  (1<<0)

#define PATTERN_ZONE_RIGHT (1<<1)

Regards,

Anil

[John E KVAM]

After we did that project we ran into some corner cases. Basically that demo works fine - for that door.

We did try hard to fix it, but that just led to more issues.

The right answer is to use a better sensor. The VL53L5CX is a multizone sensor that takes multiple zones at the same time.

I propose using this project and the VL53L5 as a solution to your problems.

It just works better.

• john

[Akuma.6]

Hi John,

You mean to say the below video link is related to the VL53L1X sensor, but the solution is not stable right?

I am trying to post the Emails to the ST team for the beginning (6 months), but nobody supports in India.

Please could somebody share the emails to address my problem. Because based on your video we designed the initial board based on VL53L1X sensor and implementing the ppl counting logic. But some portion is not recognizing and not exact count measuring using this sensor.

Your better answer is I need to migrate from VL53L1X to VL53L5X right? Will there be any software logic difference to both sensors?

Before migrating to another series, Please share the bugs or dependencies related to VL53L1X ?

Regards,

Kumar

[John E KVAM]

The VL53L1X has a theroritical range of 4 meters. But this range is dependent on getting enough photons back to be counted. So while it can see a large white wall at 4m it cannot see people beyond about 2 1/2 or 3m depending on what they are wearing. (Not the color, but the fabric itself. The dye is transparent to 940nm light.)

But if you are looking down, the cross-section of the person's body is small, and if they are short it's far away.

The next problem is that the people counting divides the receive array into 2 halves. Each range uses half the array, making the 2 zones. But that further cuts down ability to gather photons.

I was in that video, and I can claim it worked. For that door, with those people, in that condition.

But in real life meaning all doors with all people, that concept just wasn't good enough. I suspect that is what you are seeing. You detect most of the people, most of the time.

When engineers complained, sort of like you are doing now, we tried to fix it, but all we got was more detects more of the time. But we did not get all detects all the time.

ST is building better and better sensors all the time. And I suspect the only reliable way to solve this really demanding problem is to use a sensor that takes all the zones in one shot, and has more laser power.

The VL53L5 solves the multi-zone problem and it will work a lot better. But even then there are going to be corner cases. The right solution is the VL53L8 - which is due out in the next quarter. It has multiple zones, and brighter light.

I'm really sorry you invested in a system that was not up to your expectations.

Have a look at the datasheet for the VL53L8CX available on ST.com and when the evaluation kit is out, see if it works for you. Then try again.

But unfortunately both the hardware and the software has to change.