Filters

Depending on the used device, it may have different filter functions.

 

The filter...

• deletes faulty targets in the near range
• eliminates measurement errors for small pulses
• filters fog up to a certain grade, but than it becomes visible

The fog filter reduces the sensitivity of the sensor in the near field, so that fog will not be seen.

 

not implemented

The fog filter has a hardware and a software component.
The hardware component is the internal trigger level in the near range up to 12m. If the fog filter is active, this trigger level for incoming pulses is higher than normal, that means the incoming pulse intensity must be higher to be recognized as a echo.
The software component is a small pulse supression, that means small pulses (pulse length lower than 6% of the normal one) are filtered out; of course you loose one echo then, but it is an unrelevant one of course.
Both components are working on the incoming analog signal from the laser echo. It is long before the measurement data are created and packed into a string.
Scanning into fog is like driving car with travelling light into fog.

 

The fog filter in the LMS5xx is a SW filter which is adjustable.
It is effective up to a distance of 16m.
In the settings there is a slider and that means:
The slider from 1-6 defines different sensitivity levels (6 is the hardest filter, 1 is the weakest filter)
The standard setting is level 3, that is a good mean value to start outdoor and also is used most times.
If you have anyway problems with fog, than turn the slider to the right side, the Level increases then up to a possible value of 6.
As the sensitivity of the sensor is reduced than, also black targets become detected worse.
If you have not so much fog and the detectivity on black should be increased, turn the slider to the left side and the sensitivity of the sensor is increased.
That makes the detection of black targets better.

The quest here is always to find the right middle way for each application.

 

 

 

 

The fog filter works like a low pass filter curve. Nearby the scanner the light is more muted than in higher distances.Like you know it from the car, if it is foggy and you switch on the upper beam headlights you will only see a "wall" of fog. Same principle by the LMS, therefore we muted it in nearer distances.
In the LMS2xx was a fix hardware filter built in, in the LMS5xx is a software filter used. Therefore you could adjust it with the sensitivity slider in SOPAS. (since FW V1.20)

 

 

 

 

* Both have the same function
* known but unimportant targets in a certain range can be filtered out.
* Can also be used for parametrizing a minimum range
* only complete signals can be received.
The device closes the receiver diode for the time frame the light needs from sending out the light pulse until it comes back from the parametrized range. But as the sent out pulse has a certain length, it makes no sense to parametrize the hardware gating to a lower value than 2m, because below 2m, not the complete pulse can be received and so it will not be evaluated.
The hardware gating also may delete a first back comming pulse because the sensors "Eye" is closed, so in some cases the second pulse is essential for getting a measurement value.

 

 

 

 

* filters out single small irrelevant targets
e.g. dust particles, rain, snow, but it depends on the amount of particles how much can be filtered out
The Particle filter considers for the actual made measurement point, if it is plausible in the actual environment.
It takes into account the measurement value from one scan before and after the actual value. That is why we get one scan delay in the evaluation, if the particle filter is active.
It also evaluates the angular degree before and after the actual measurement Point.
If the actual measured value does not fit into the environment of these four neighbour measurements, the sensor replaces the "implausible" value of the current scan by the value of the same angle in the previous scan.

An example is like the following:
If the100th point of the No. n scan is judged as "implausible", it will be replaced by the 100th point of No. (n-1) scan.  Note: No. (n-1) scan is the previous scan of No. n scan.

Additional hint: for measurement application under the fast changing/moving scenario, maybe it is better to turn off "particle filter", otherwise a measurement delay of one scan may happen. Also it could be, that small objects which you want to see, but they are one spot wide only , will be deleted!

 

 

-The particle filter is created for elimination of small particles that hit one spot.
-When activated, the scanner always evaluates three scans: The actual scan, the scan before and the scan after. This is why the data output is shifted by one scan: You will never get the real actual one, but always the one before the actual one.
-The scanner checks the actual angular degree, the (angular) step before and the step after and than it checks if the actual value fits in that area.
-In the example, the actual value is "5". The scanner checks the values around and recognizes that all the other values are in the "36 - 45" range => "5" does not fit in there. This is why the scanner replaces that actual value by the value it had in the same scan but in the angular step before. Maybe that is a filtered value too, but it will be a value that fits in the environment.
-The filter calculation is not including the angular steps before and after the actual one (green line), because it might happen that a particle hits two consecutive overlapping spots. In that case, the scanner will have the (one) particle divided on two spots and considers the particle as real object (because it is bigger than one spot).
-The calculation is also considering the measured distances between the calculated spots. The distance of the actual calculated value must be bigger than 0.5m to the surrounding values to be considered as a particle, otherwise the scanner does not "use" the filter.
 

 

 

 

* If parametrized, the first pulses were filtered out, only the last is evaluated
* between the first and the second pulse, there has to be a distance of min. 1m, otherwise the second pulse will be deleted. That means, that between the two seen targets, there has to be a distance of at least 1m, otherwise the second one will not be seen. For example, an object which has to be detected behind a window (and the window itself creates an echo too), has to be at minimum 1m behind the window.
For the measurement data that means:
If the filter is active (and both or all five channels are given out), in the second echo, every measurement value which was there is filtered out by a filter (value = 3 means filtered value)
The first scan in the example below is with filter inactive, second with filter active, third with filter inactive, fourth with filter active.
So If the filter is active, only one(Channel 1) measurement data channel will be there, the other one (channel 2) is filtered out, because only the last echo is taken and put into channel 1.

In LMS1xx former called: n-Pulse to 1-Pulse Filter

 

The Meanfilter filters the measurement data only for the output on communication interfaces and will not effects the data processed within the field evaluation.
Every spot data is averaged over "Number of scans" values. After "Number of scans" the result will put on the communication interface, so that the data output rate is divided by "Number of scans".
Invalid values will not be considered in the calculation, i.e.:
* "Number of scans" = 10
* 3 values within the same angle are invalid
→  Mean value = (Sum of valid distances) / 7

The edge filter prevents incorrect or extreme distance values at edges. These are caused by laser light that partially hits an object in the foreground and partially hits an object further away or by too little light emission at the actual object (see "Influences of object surfaces on the measurement"). The filter evaluates the distance difference between neighboring points.


The preceding and following measuring points are considered in each case. As soon as the parameterized "maximum neighboring distance" to the previous and subsequent measuring point is exceeded, the device sets the distance value to 0 (values in table x highlighted in bold). If the "maximum neighboring distance" is not exceeded or is only exceeded to one of the two neighboring measuring points, then the measuring point is not filtered. If the measuring point or one of the neighboring measuring points is equal to 0, then the filter is not applied to this measuring point.
 

 

 

Maximum edge filter range (range limitation).

The maximum distance can be set up to 8000 mm. The measuring points within the parameterized range are taken into calculation.

Because of the bigger measurement range the LMS14x is more sensitive to noise. So the measurement might be affect by direct sunlight. This situation is detected and the according scan beams are marked as dazzled. Depending on the settings evaluation cases might consider dazzled beams as field violation. If this is a problem for an application the dazzling filter can be used. The filter will flag any dazzled beam as "filtered" instead. The field evaluation reacts differently to filtered beams so the customer might improve the availability of his application. Enabling the filter will leave this device more vulnerable to sabotage though.

 

The LMS5xx has dazzling filter from FW 2.10

 

The sensor takes three consequtive measurement points and sorts them  by highest value.
Than he takes the middle one and gives it out as a 1° measurement value.
The data output rate is than reduced to 1°, but the measurement itself is still 0,33°

The sliding average filter smooths the distance value. It does this by calculating the arithmetic mean from several scans of the same point. The number of scans can be configured (maximum 4 scans).

Individual outliers (shown in bold in the table) influence the average value. Once the measured value telegram has been confirmed, the first measured value is not output until after the configured number of scans. Therefore, there is always a time delay equivalent to the number of scans configured for averaging. The digit of the first scan included in the averaging calculation is always output in the scan counter. Invalid distance values (= 0) are not included in the averaging calculation, so that in these places a smaller number of scans is used in the division calculation. Based on a scanning frequency of 15 Hz, a measured value is generated every 67 ms . The time delay affecting data output results from this base value multiplied by the number of averaging operations (e.g., 2 averagings = 134 ms, 4 averagings =268ms).