Polyline encode gets wrong lat/lng after decoding - google-polyline

We are using Google's Polyline decoding algorithm to decode our coordinates. But in our case the most coordinates are wrong after decoding it. We have also tested the process with a deeper precision.
This is our code and also our logs to test that the coordinates are wrong:
let coordinates = [ [lat, lng], [...], ...];
console.log(coordinates[13347]); // Output: [ 13.44668, 52.47429 ]
let encoded = Polyline.encode(coordinates);
let decoded = Polyline.decode(encoded);
console.log(decoded[13347]); // Output: [ 13.44671, 52.47445 ]
console.log(coordinates.length == decoded.length)// true
In this case the distance is 20 meters which is a lot. Other points have distances like 150 meter or even more.
In my coordinates array are around 250.000 coordinates which we want to decode.
Am I missing something so the decode/encode process fails this hard ?

TL;DR
Add the following lines after the declaration of the coordinates variable:
coordinates = coordinates.map(
pair => { return [pair[0].toFixed(5), pair[1].toFixed(5)]; }
);
Full answer
It looks like you're dealing with floating point rounding errors. Probably the library you use has incorrect implementation of the Polyline encoding algorithm.
In the description of the algorithm we read that the encoded string generated by the algorithm stores the differences between consecutive coordinates using fixed-precision numbers (with 5 decimal places). Therefore it is important to round the latitude and longitude to 5 decimal places before computing the differences. Without that step, the rounding errors may accumulate. In the worst case error may increase by about 0.000005 deg for each subsequent item in the encoded list.
The official implementation of the algorithm does not introduce accumulated rounding errors. However, the implementation found in NPM (package polyline) gives incorrect results that indicate the invalid rounding of numbers.
Please look at the examples bellow:
Example 1. Encoding a polyline using official implementation of the algorithm
(using google.maps.geometry.encoding.encodePath from the Google Maps JavaScript API)
originalList = [];
for (var i = 0; i < 100; ++i)
originalList.push(
new google.maps.LatLng(6 * i / 1000000, 0)
);
// originalList looks like: [[0.000000,0],[0.000006,0],[0.000012,0],[0.000018,0], ..., [0.000594,0]];
// (but with LatLng objects instead of 2-element arrays)
console.log(originalList[99].lat()) // 0.000594
var encodedList = google.maps.geometry.encoding.encodePath(originalList)
var decodedList = google.maps.geometry.encoding.decodePath(encodedList)
console.log(decodedList[99].lat()) // 0.00059
Example 2. Encoding a polyline using package polyline from NPM
let Polyline = require('polyline');
var originalList = [];
for (var i = 0; i < 100; ++i)
originalList.push(
[6 * i / 1000000, 0]
);
// again: originalList == [[0.000000,0],[0.000006,0],[0.000012,0],[0.000018,0], ..., [0.000594,0]];
console.log(originalList[99][0]) // 0.000594
var encodedList = Polyline.encode(originalList);
var decodedList = Polyline.decode(encodedList);
console.log(decodedList[99][0]) // 0.00099
Invalid result: the values 0.000594 and 0.00099 differ by more than 0.000005.
Possible fix
The library that you're using probably doesn't round the coordinates before computing the differences.
For example when two consecutive points have latitudes 0.000000 and 0.000006, the difference is 0.000006 and it is rounded to 0.00001 giving error of 0.000004.
You may want to round the coordinates manually, before passing them to Polyline.encode(), eg. using the function .toFixed(5):
let Polyline = require('polyline');
var originalList = [];
for (var i = 0; i < 100; ++i)
originalList.push(
[(6 * i / 1000000).toFixed(5), 0]
);
// before rounding: [[ 0.000000,0],[ 0.000006,0],[ 0.000012,0],[ 0.000018,0], ..., [ 0.000594,0]];
// after rounding: [['0.00000',0],['0.00001',0],['0.00001',0],['0.00002',0], ..., ['0.00059',0]];
console.log(originalList[99][0]) // 0.00059
var encodedList = Polyline.encode(originalList);
var decodedList = Polyline.decode(encodedList);
console.log(decodedList[99][0]) // 0.00059

Polyline encoding is lossy:
https://developers.google.com/maps/documentation/utilities/polylinealgorithm (Polyline encoding is a *lossy* compression algorithm that allows you to store a series of coordinates as a single string
How about using your own encoding scheme? The page above also shows the encoding scheme used by Google. Perhaps you can look for a trade-off between space and accuracy.

Related

creating audio file based on frequencies

I'm using node.js for a project im doing.
The project is to convert words into numbers and then to take those numbers and create an audio output.
The audio output should play the numbers as frequencies. for example, I have an array of numbers [913, 250,352] now I want to play those numbers as frequencies.
I know I can play them in the browser with audio API or any other third package that allows me to do so.
The thing is that I want to create some audio file, I tried to convert those numbers into notes and then save it as Midi file, I succeed but the problem is that the midi file takes the frequencies, convert them into the closest note (example: 913 will convert into 932.33HZ - which is note number 81),
// add a track
var array = gematriaArray
var count = 0
var track = midi.addTrack()
var note
for (var i = 0; i < array.length; i++) {
note = array[i]
track = track.addNote({
//here im converting the freq -> midi note.
midi: ftom(parseInt(note)),
time: count,
duration: 3
})
count++
}
// write the output
fs.writeFileSync('./public/sounds/' + name + random + '.mid', new Buffer.from(midi.toArray()))
I searched the internet but I couldn't find anything that can help.
I really want to have a file that the user can download with those numbers as frequencies, someone knows what can be done to get this result?
Thanks in advance for the helpers.
this function will populate a buffer with floating point values which represent the height of the raw audio curve for the given frequency
var pop_audio_buffer_custom = function (number_of_samples, given_freq, samples_per_second) {
var number_of_samples = Math.round(number_of_samples);
var audio_obj = {};
var source_buffer = new Float32Array(number_of_samples);
audio_obj.buffer = source_buffer;
var incr_theta = (2.0 * Math.PI * given_freq) / samples_per_second;
var theta = 0.0;
for (var curr_sample = 0; curr_sample < number_of_samples; curr_sample++) {
audio_obj.buffer[curr_sample] = Math.sin(theta);
console.log(audio_obj.buffer[curr_sample] , "theta ", theta);
theta += incr_theta;
}
return audio_obj;
}; // pop_audio_buffer_custom
var number_of_samples = 10000; // long enough to be audible
var given_freq = 300;
var samples_per_second = 44100; // CD quality sample rate
var wav_output_filename = "/tmp/wav_output_filename.wav"
var synthesized_obj = {};
synthesized_obj.buffer = pop_audio_buffer_custom(number_of_samples, given_freq, samples_per_second);
the world of digital audio is non trivial ... the next step once you have an audio buffer is to translate the floating point representation into something which can be stored in bytes ( typically 16 bit integers dependent on your choice of bit depth ) ... then that 16 bit integer buffer needs to get written out as a WAV file
audio is a wave sometimes called a time series ... when you pound your fist onto the table the table wobbles up and down which pushes tiny air molecules in unison with that wobble ... this wobbling of air propagates across the room and reaches a microphone diaphragm or maybe your eardrum which in turn wobbles in resonance with this wave ... if you glued a pencil onto the diaphragm so it wobbled along with the diaphragm and you slowly slid a strip of paper along the lead tip of the pencil you would see a curve being written onto that paper strip ... this is the audio curve ... an audio sample is just the height of that curve at an instant of time ... if you repeatedly wrote down this curve height value X times per second at a constant rate you will have a list of data points of raw audio ( this is what above function creates ) ... so a given audio sample is simply the value of the audio curve height at a given instant in time ... since computers are not continuous instead are discrete they cannot handle the entire pencil drawn curve so only care about this list of instantaneously measured curve height values ... those are audio samples
above 32 bit floating point buffer can be fed into following function to return a 16 bit integer buffer
var convert_32_bit_float_into_signed_16_bit_int_lossy = function(input_32_bit_buffer) {
// this method is LOSSY - intended as preliminary step when saving audio into WAV format files
// output is a byte array where the 16 bit output format
// is spread across two bytes in little endian ordering
var size_source_buffer = input_32_bit_buffer.length;
var buffer_byte_array = new Int16Array(size_source_buffer * 2); // Int8Array 8-bit twos complement signed integer
var value_16_bit_signed_int;
var index_byte = 0;
console.log("size_source_buffer", size_source_buffer);
for (var index = 0; index < size_source_buffer; index++) {
value_16_bit_signed_int = ~~((0 < input_32_bit_buffer[index]) ? input_32_bit_buffer[index] * 0x7FFF :
input_32_bit_buffer[index] * 0x8000);
buffer_byte_array[index_byte] = value_16_bit_signed_int & 0xFF; // bitwise AND operation to pluck out only the least significant byte
var byte_two_of_two = (value_16_bit_signed_int >> 8); // bit shift down to access the most significant byte
buffer_byte_array[index_byte + 1] = byte_two_of_two;
index_byte += 2;
};
// ---
return buffer_byte_array;
};
next step is to persist above 16 bit int buffer into a wav file ... I suggest you use one of the many nodejs libraries for that ( or even better write your own as its only two pages of code ;-)))

How to get a consistent hash/spectra of n points given a 32-byte md5 string

I would like to take an md5 hash of some content, and then generate a "curve" or a "spectrum" so to speak, of n points. That is, to plot let's say 5, 10, or 20 points on a line from 0 to 1, distributed in a way so that it's unique to the md5 hash (collisions don't much matter). Basically it would look like a atomic light emission spectrum.
These points (or lines in the spectra) are somehow generated based on the md5 hash provided, and the n provided saying how many lines you want.
So it would be like:
function generateSpecrum(md5, n) { return [ ... ] }
By default it could just return the values from between 0 and 1, but maybe you give it a start and end value from which to generate the range.
Wondering how this could be done, in pseudocode or in JS.
There will be however many possibilities of a standard md5 hash. I would just do this:
var crypto = require('crypto')
var data = 'foo'
crypto.createHash('md5').update(data).digest('hex')
// acbd18db4cc2f85cedef654fccc4a4d8
So a 32-byte string. In my case it doesn't need to produce globally unique values, there can be some collisions, but if there was a way for it to produce a variety of spectra from different md5 inputs, that would be cool.
Let's ignore the part where the string data is an md5 print and instead focus on how to do this for arbitrary length hexadecimal strings, so we can use any digest we like (from CRC32 to SHA-512):
start with a hue gradient background (we can do this in CSS),
turn the string into a bit print (built into JS), and
black out any region that corresponds to a zero bit.
As a runnable snippet:
function hexstr2bin(stringinput) {
// let's not be constrained by JS integer precision,
// which is only good for 53 bits. Technically we don't
// care what the "numbers" are here, we just want the
// ones and zeros that the numbers turn into.
return stringinput.split('').map(c => (
parseInt(c, 16).toString(2).padStart(4,'0')
)).join('');
}
function renderSpectrum(stringinput) {
let cvs = document.createElement('canvas');
let bits = Array.from(hexstr2bin(stringinput));
cvs.width = bits.length;
cvs.height = 1;
ctx = cvs.getContext('2d');
ctx.strokeStyle = 'black';
bits.forEach( (bit,i) => {
if (bit === "0") {
ctx.moveTo(i,0);
ctx.lineTo(i,1);
ctx.stroke();
}
});
document.body.appendChild(cvs);
};
renderSpectrum("acbd18db4fccc4a4d8");
renderSpectrum("c5887c91d0002f2a869a4b0772827701");
renderSpectrum("06956ff032d78e090d0d292aa9d8e7143ab08cf1ed444944529f79a4f937306a");
canvas {
width: 100%;
height: 40px;
background: linear-gradient(
to right,
violet, blue, cyan, green, yellow, orange, red
);
}
And stretching the canvas to 100% width means you get blurring for free. Bonus!

Custom filter bank is not generating the expected output

Please, refer to this article.
I have implemented the section 4.1 (Pre-processing).
The preprocessing step aims to enhance image features along a set of
chosen directions. First, image is grey-scaled and filtered with a
sharpening filter (we subtract from the image its local-mean filtered
version), thus eliminating the DC component.
We selected 12 not overlapping filters, to analyze 12 different
directions, rotated with respect to 15° each other.
GitHub Repositiry is here.
Since, the given formula in the article is incorrect, I have tried two sets of different formulas.
The first set of formula,
The second set of formula,
The expected output should be,
Neither of them are giving proper results.
Can anyone suggest me any modification?
GitHub Repository is here.
Most relevalt part of the source code is here:
public List<Bitmap> Apply(Bitmap bitmap)
{
Kernels = new List<KassWitkinKernel>();
double degrees = FilterAngle;
KassWitkinKernel kernel;
for (int i = 0; i < NoOfFilters; i++)
{
kernel = new KassWitkinKernel();
kernel.Width = KernelDimension;
kernel.Height = KernelDimension;
kernel.CenterX = (kernel.Width) / 2;
kernel.CenterY = (kernel.Height) / 2;
kernel.Du = 2;
kernel.Dv = 2;
kernel.ThetaInRadian = Tools.DegreeToRadian(degrees);
kernel.Compute();
//SleuthEye
kernel.Pad(kernel.Width, kernel.Height, WidthWithPadding, HeightWithPadding);
Kernels.Add(kernel);
degrees += degrees;
}
List<Bitmap> list = new List<Bitmap>();
Bitmap image = (Bitmap)bitmap.Clone();
//PictureBoxForm f = new PictureBoxForm(image);
//f.ShowDialog();
Complex[,] cImagePadded = ImageDataConverter.ToComplex(image);
Complex[,] fftImage = FourierTransform.ForwardFFT(cImagePadded);
foreach (KassWitkinKernel k in Kernels)
{
Complex[,] cKernelPadded = k.ToComplexPadded();
Complex[,] convolved = Convolution.ConvolveInFrequencyDomain(fftImage, cKernelPadded);
Bitmap temp = ImageDataConverter.ToBitmap(convolved);
list.Add(temp);
}
return list;
}
Perhaps the first thing that should be mentioned is that the filters should be generated with angles which should increase in FilterAngle (in your case 15 degrees) increments. This can be accomplished by modifying KassWitkinFilterBank.Apply as follow (see this commit):
public List<Bitmap> Apply(Bitmap bitmap)
{
// ...
// The generated template filter from the equations gives a line at 45 degrees.
// To get the filter to highlight lines starting with an angle of 90 degrees
// we should start with an additional 45 degrees offset.
double degrees = 45;
KassWitkinKernel kernel;
for (int i = 0; i < NoOfFilters; i++)
{
// ... setup filter (unchanged)
// Now increment the angle by FilterAngle
// (not "+= degrees" which doubles the value at each step)
degrees += FilterAngle;
}
This should give you the following result:
It is not quite the result from the paper and the differences between the images are still quite subtle, but you should be able to notice that the scratch line is most intense in the 8th figure (as would be expected since the scratch angle is approximately 100-105 degrees).
To improve the result, we should feed the filters with a pre-processed image in the same way as described in the paper:
First, image is grey-scaled and filtered with a sharpening filter (we subtract from the image its local-mean filtered version), thus eliminating the DC component
When you do so, you will get a matrix of values, some of which will be negative. As a result this intermediate processing result is not suitable to be stored as a Bitmap. As a general rule when performing image processing, you should keep all intermediate results in double or Complex as appropriate, and only convert back the final result to Bitmap for visualization.
Integrating your changes to add image sharpening from your GitHub repository while keeping intermediate results as doubles can be achieve by changing the input bitmap and temporary image variables to use double[,] datatype instead of Bitmap in the KassWitkinFilterBank.Apply method (see this commit):
public List<Bitmap> Apply(double[,] bitmap)
{
// [...]
double[,] image = (double[,])bitmap.Clone();
// [...]
}
which should give you the following result:
Or to better highlight the difference, here is figure 1 (0 degrees) on the left, next to figure 8 (105 degrees) on the right:

How to detect string tone from FFT

I've got spectrum from a Fourier transformation. It looks like this:
Police was just passing nearby
Color represents intensity.
X axis is time.
Y axis is frequency - where 0 is at top.
While whistling or a police siren leave only one trace, many other tones seem to contain a lot of harmonic frequencies.
Electric guitar plugged directly into microphone (standard tuning)
The really bad thing is, that as you can see there is no major intensity - there are 2-3 frequencies that are almost equal.
I have written a peak detection algorithm to highlight the most sigificant peak:
function findPeaks(data, look_range, minimal_val) {
if(look_range==null)
look_range = 10;
if(minimal_val == null)
minimal_val = 20;
//Array of peaks
var peaks = [];
//Currently the max value (that might or might not end up in peaks array)
var max_value = 0;
var max_value_pos = 0;
//How many values did we check without changing the max value
var smaller_values = 0;
//Tmp variable for performance
var val;
var lastval=Math.round(data.averageValues(0,4));
//console.log(lastval);
for(var i=0, l=data.length; i<l; i++) {
//Remember the value for performance and readibility
val = data[i];
//If last max value is larger then the current one, proceed and remember
if(max_value>val) {
//iterate the ammount of values that are smaller than our champion
smaller_values++;
//If there has been enough smaller values we take this one for confirmed peak
if(smaller_values > look_range) {
//Remember peak
peaks.push(max_value_pos);
//Reset other variables
max_value = 0;
max_value_pos = 0;
smaller_values = 0;
}
}
//Only take values when the difference is positive (next value is larger)
//Also aonly take values that are larger than minimum thresold
else if(val>lastval && val>minimal_val) {
//Remeber this as our new champion
max_value = val;
max_value_pos = i;
smaller_values = 0;
//console.log("Max value: ", max_value);
}
//Remember this value for next iteration
lastval = val;
}
//Sort peaks so that the largest one is first
peaks.sort(function(a, b) {return -data[a]+data[b];});
//if(peaks.length>0)
// console.log(peaks);
//Return array
return peaks;
}
The idea is, that I walk through the data and remember a value that is larger than thresold minimal_val. If the next look_range values are smaller than the chosen value, it's considered peak. This algorithm is not very smart but it's very easy to implement.
However, it can't tell which is the major frequency of the string, much like I anticipated:
The red dots highlight the strongest peak
Here's a jsFiddle to see how it really works (or rather doesn't work).
What you see in the spectrum of a string tone is the set of harmonics at
f0, 2*f0, 3*f0, ...
with f0 being the fundamental frequency or pitch of your string tone.
To estimate f0 from the spectrum (Output of FFT, abs value, probably logarithmic) you should not look for the strongest component, but the distance between all these harmonics.
One very nice method to do so is a second (inverse) FFT of the (abs, real) spectrum. This produces a strong line at t0 == 1/f0.
The sequence fft -> abs() -> fft-1 is equivalent to calculating the auto-correlation function (ACF) thanks to the Wiener–Khinchin theorem.
The precission of this approach depends on the length of the FFT (or ACF) and your sampling rate. You can improve precission a lot if you interpolate the "real" max between the sampling points of the result using a sinc function.
For even better results you could correct the intermediate spectrum: Most sounds have an average pink spectrum. If you amplify the higher frequencies (according an inverse pink spectrum) before the inverse FFT the ACF will be "better" (It takes the higher harmonics more into account, improving acuracy).

How to compute the visible area based on a heightmap?

I have a heightmap. I want to efficiently compute which tiles in it are visible from an eye at any given location and height.
This paper suggests that heightmaps outperform turning the terrain into some kind of mesh, but they sample the grid using Bresenhams.
If I were to adopt that, I'd have to do a line-of-sight Bresenham's line for each and every tile on the map. It occurs to me that it ought to be possible to reuse most of the calculations and compute the heightmap in a single pass if you fill outwards away from the eye - a scanline fill kind of approach perhaps?
But the logic escapes me. What would the logic be?
Here is a heightmap with a the visibility from a particular vantagepoint (green cube) ("viewshed" as in "watershed"?) painted over it:
Here is the O(n) sweep that I came up with; I seems the same as that given in the paper in the answer below How to compute the visible area based on a heightmap? Franklin and Ray's method, only in this case I am walking from eye outwards instead of walking the perimeter doing a bresenhams towards the centre; to my mind, my approach would have much better caching behaviour - i.e. be faster - and use less memory since it doesn't have to track the vector for each tile, only remember a scanline's worth:
typedef std::vector<float> visbuf_t;
inline void map::_visibility_scan(const visbuf_t& in,visbuf_t& out,const vec_t& eye,int start_x,int stop_x,int y,int prev_y) {
const int xdir = (start_x < stop_x)? 1: -1;
for(int x=start_x; x!=stop_x; x+=xdir) {
const int x_diff = abs(eye.x-x), y_diff = abs(eye.z-y);
const bool horiz = (x_diff >= y_diff);
const int x_step = horiz? 1: x_diff/y_diff;
const int in_x = x-x_step*xdir; // where in the in buffer would we get the inner value?
const float outer_d = vec2_t(x,y).distance(vec2_t(eye.x,eye.z));
const float inner_d = vec2_t(in_x,horiz? y: prev_y).distance(vec2_t(eye.x,eye.z));
const float inner = (horiz? out: in).at(in_x)*(outer_d/inner_d); // get the inner value, scaling by distance
const float outer = height_at(x,y)-eye.y; // height we are at right now in the map, eye-relative
if(inner <= outer) {
out.at(x) = outer;
vis.at(y*width+x) = VISIBLE;
} else {
out.at(x) = inner;
vis.at(y*width+x) = NOT_VISIBLE;
}
}
}
void map::visibility_add(const vec_t& eye) {
const float BASE = -10000; // represents a downward vector that would always be visible
visbuf_t scan_0, scan_out, scan_in;
scan_0.resize(width);
vis[eye.z*width+eye.x-1] = vis[eye.z*width+eye.x] = vis[eye.z*width+eye.x+1] = VISIBLE;
scan_0.at(eye.x) = BASE;
scan_0.at(eye.x-1) = BASE;
scan_0.at(eye.x+1) = BASE;
_visibility_scan(scan_0,scan_0,eye,eye.x+2,width,eye.z,eye.z);
_visibility_scan(scan_0,scan_0,eye,eye.x-2,-1,eye.z,eye.z);
scan_out = scan_0;
for(int y=eye.z+1; y<height; y++) {
scan_in = scan_out;
_visibility_scan(scan_in,scan_out,eye,eye.x,-1,y,y-1);
_visibility_scan(scan_in,scan_out,eye,eye.x,width,y,y-1);
}
scan_out = scan_0;
for(int y=eye.z-1; y>=0; y--) {
scan_in = scan_out;
_visibility_scan(scan_in,scan_out,eye,eye.x,-1,y,y+1);
_visibility_scan(scan_in,scan_out,eye,eye.x,width,y,y+1);
}
}
Is it a valid approach?
it is using centre-points rather than looking at the slope between the 'inner' pixel and its neighbour on the side that the LoS passes
could the trig in to scale the vectors and such be replaced by factor multiplication?
it could use an array of bytes since the heights are themselves bytes
its not a radial sweep, its doing a whole scanline at a time but away from the point; it only uses only a couple of scanlines-worth of additional memory which is neat
if it works, you could imagine that you could distribute it nicely using a radial sweep of blocks; you have to compute the centre-most tile first, but then you can distribute all immediately adjacent tiles from that (they just need to be given the edge-most intermediate values) and then in turn more and more parallelism.
So how to most efficiently calculate this viewshed?
What you want is called a sweep algorithm. Basically you cast rays (Bresenham's) to each of the perimeter cells, but keep track of the horizon as you go and mark any cells you pass on the way as being visible or invisible (and update the ray's horizon if visible). This gets you down from the O(n^3) of the naive approach (testing each cell of an nxn DEM individually) to O(n^2).
More detailed description of the algorithm in section 5.1 of this paper (which you might also find interesting for other reasons if you aspire to work with really enormous heightmaps).

Resources