100,000 pictures, how to find two similar pictures?

100,000 photos, how to find two similar pictures??? Urgent and urgent

The company has a project, need to protect customer's design information, customer information are pictures, how to find two similar pictures from 10w, these two pictures from the same camera shooting, is more like, like face recognition technology, how to achieve similar, than must be similar to two, Can be a similar number of pictures, later we will be artificial comparison?

------to solve the idea----------------------

/**
* Generate fingerprints
* $filename picture file name, can be URL. Only the type of image supported by GD
* $k sampled number of dots
* $retmode return Format: 2 binary represents 1 array 26 binary representation
**/
function Producefingerprint ($filename, $k =8, $retmode =0) {
$sim = imagecreatefromstring (file_get_contents ($filename));
$dim = Imagecreate ($k, $k);
Imagecopyresized ($dim, $sim, 0, 0, 0, 0, $k, $k, Imagesx ($sim), Imagesy ($sim));
Imagetruecolortopalette ($sim, True, 64);
ImageFilter ($sim, Img_filter_grayscale);

for ($x =0; $x < $k; $x + +) for ($y =0; $y < $k; $y + +) {
$c = Imagecolorat ($dim, $x, $y);
$p [] = Array_sum (Imagecolorsforindex ($dim, $c))/3;
}
$avl = Array_sum ($p)/count ($p);
$r = ";
foreach ($p as $v) $r. = $v >= $avl? 1:0;

if ($retmode = = 0) return $r;
$p = Array_map (' Bindec ', Str_split ($r, 8));
if ($retmode = = 1) return $p;
if ($retmode = = 2) {
Array_unshift ($p, '%02x%02x%02x%02x%02x%02x%02x%02x ');
Return Call_user_func_array (' sprintf ', $p);
}
return false;
}

Use this function to generate a fingerprint for each picture, save to the database backup

When checking, the fingerprint is the same as the picture, which does not have to say
Calculation of similarity

$p 1 = producefingerprint ($f 1, $k);
$p 2 = Producefingerprint ($f 2, $k);
$n = Count (Array_diff_assoc (Str_split ($p 1), Str_split ($p 2));
Echo 1-$n/strlen ($p 1);

------to solve the idea----------------------
Similarity (a value less than or equal to 1), the threshold is up to you
Generally think
0.95 very Similar
0.9 Similar
0.85 similar
It's not like it's a little.
------to solve the idea----------------------
Similarity I have tested, photo shooting angle, light and shade on the similarity effect is great!

