identify objects more precisely

svn path=/trunk/externals/pdp_opencv/; revision=11623

2 files changed, 174 insertions, 56 deletions

diff --git a/pdp_opencv_haarcascade-help.pd b/pdp_opencv_haarcascade-help.pd
index 4fc5ed1..af0cbf9 100644
--- a/pdp_opencv_haarcascade-help.pd
+++ b/pdp_opencv_haarcascade-help.pd
@@ -1,4 +1,4 @@
-#N canvas 25 49 1247 801 10;
+#N canvas 30 49 1247 801 10;
 #X obj -158 12 cnv 15 621 250 empty empty empty 20 12 0 14 -260097
 -66577 0;
 #X obj 82 52 tgl 15 0 empty empty empty 17 7 0 10 -262144 -1 -1 0 1
@@ -29,59 +29,68 @@
 #X obj -14 108 loadbang;
 #X text -85 41 playing a video file;
 #X text 112 50 Camera input;
-#X obj -156 266 cnv 15 621 300 empty empty empty 20 12 0 14 -258113
+#X obj -158 272 cnv 15 621 350 empty empty empty 20 12 0 14 -258113
 -66577 0;
-#X obj -129 591 pdp_xv;
-#X floatatom 25 511 5 0 0 0 - - -;
-#X floatatom 74 512 5 0 0 0 - - -;
-#X floatatom 121 510 5 0 0 0 - - -;
-#X obj 21 465 route 0 1 2 3 4;
-#X obj -129 418 pdp_opencv_haarcascade;
-#X text 123 469 For each object detected;
-#X obj 24 489 unpack 0 0 0;
-#X floatatom 174 512 5 0 0 0 - - -;
-#X floatatom 213 512 5 0 0 0 - - -;
-#X floatatom 251 513 5 0 0 0 - - -;
-#X obj 172 491 unpack 0 0 0;
+#X obj -130 661 pdp_xv;
+#X floatatom 25 583 5 0 0 0 - - -;
+#X floatatom 74 584 5 0 0 0 - - -;
+#X floatatom 121 582 5 0 0 0 - - -;
+#X obj 21 537 route 0 1 2 3 4;
+#X obj -129 490 pdp_opencv_haarcascade;
+#X text 123 541 For each object detected;
+#X obj 24 561 unpack 0 0 0;
+#X floatatom 174 584 5 0 0 0 - - -;
+#X floatatom 213 584 5 0 0 0 - - -;
+#X floatatom 251 585 5 0 0 0 - - -;
+#X obj 172 563 unpack 0 0 0;
 #X text -158 -62 pdp_opencv_haarscasccade : object recognition based
 on Haar's algorithm ( decision tree algorithm );
 #X text -159 -37 used for face detection by default;
-#X floatatom -56 443 5 0 0 0 - - -;
-#X text -17 443 number of objects detected;
-#X msg 6 282 load \$1;
-#X obj 70 282 openpanel;
-#X obj 138 281 bng 15 250 50 0 empty empty empty 17 7 0 10 -262144
+#X floatatom -56 515 5 0 0 0 - - -;
+#X text -17 515 number of objects detected;
+#X msg -27 309 load \$1;
+#X obj 37 309 openpanel;
+#X obj 105 308 bng 15 250 50 0 empty empty empty 17 7 0 10 -262144
 -1 -1;
 #X text -162 -25 ( load haarscascase_frontalface_alt.xml );
-#X text 161 292 ( default haarscascase_frontalface_alt.xml );
-#X text 470 288 don't forget this step;
-#X text 471 274 <-- WARNING : if you don't see any object detected
+#X text 128 319 ( default haarscascase_frontalface_alt.xml );
+#X text 471 294 don't forget this step;
+#X text 472 280 <-- WARNING : if you don't see any object detected
 ;
-#X text 163 280 load an haar's decision tree;
-#X msg 25 311 mode \$1;
-#X obj 83 311 hradio 15 1 0 2 empty empty empty 0 -8 0 10 -262144 -1
+#X text 130 307 load an haar's decision tree;
+#X msg -8 338 mode \$1;
+#X obj 50 338 hradio 15 1 0 2 empty empty empty 0 -8 0 10 -262144 -1
 -1 0;
-#X msg 34 338 min_size \$1;
-#X floatatom 114 340 5 0 0 0 - - -;
-#X floatatom 317 512 5 0 0 0 - - -;
-#X floatatom 356 512 5 0 0 0 - - -;
-#X floatatom 394 513 5 0 0 0 - - -;
-#X obj 315 491 unpack 0 0 0;
-#X text 124 311 mode 0 or 1 ( default 0 );
-#X text 154 338 minimum size of object ( default 30 );
-#X floatatom 152 363 5 0 0 0 - - -;
-#X text 190 362 minimum neighbours of object ( default 2 );
-#X msg 46 362 min_neighbors \$1;
-#X floatatom 160 389 5 0 0 0 - - -;
-#X text 198 388 scale used for recognition ( default 1.1 );
-#X text 315 527 Xcenter Ycenter Radius;
-#X text 171 527 Xcenter Ycenter Radius;
-#X text 25 527 Xcenter Ycenter Radius;
-#X msg 54 388 scale_factor \$1;
+#X msg 1 365 min_size \$1;
+#X floatatom 81 367 5 0 0 0 - - -;
+#X floatatom 317 584 5 0 0 0 - - -;
+#X floatatom 356 584 5 0 0 0 - - -;
+#X floatatom 394 585 5 0 0 0 - - -;
+#X obj 315 563 unpack 0 0 0;
+#X text 91 338 mode 0 or 1 ( default 0 );
+#X text 121 365 minimum size of object ( default 30 );
+#X floatatom 119 390 5 0 0 0 - - -;
+#X text 157 389 minimum neighbours of object ( default 2 );
+#X msg 13 389 min_neighbors \$1;
+#X floatatom 127 416 5 0 0 0 - - -;
+#X text 165 415 scale used for recognition ( default 1.1 );
+#X text 315 599 Xcenter Ycenter Radius;
+#X text 171 599 Xcenter Ycenter Radius;
+#X text 25 599 Xcenter Ycenter Radius;
+#X msg 21 415 scale_factor \$1;
 #X obj -68 228 hsl 128 15 0 500 0 0 empty empty empty -2 -8 0 10 -262144
 -1 -1 0 1;
 #X text -158 -9 Written by Lluis Gomez i Bigorda ( lluisgomez@hangar.org
 );
+#X obj -129 687 route press;
+#X msg -79 661 cursor 1;
+#X msg 40 439 ftolerance \$1;
+#X floatatom 130 440 5 0 0 0 - - -;
+#X msg -58 209 162;
+#X msg 45 463 clear;
+#X text 90 464 clear history;
+#X text 171 447 temporarily ( default 5 );
+#X text 171 435 number of frames where an object can disappear;
 #X connect 1 0 2 0;
 #X connect 2 0 4 0;
 #X connect 3 0 4 0;
@@ -105,10 +114,12 @@ on Haar's algorithm ( decision tree algorithm );
 #X connect 21 0 20 0;
 #X connect 22 0 21 0;
 #X connect 22 0 15 0;
+#X connect 26 0 71 0;
 #X connect 30 0 33 0;
 #X connect 30 1 37 0;
 #X connect 30 2 57 0;
 #X connect 31 0 26 0;
+#X connect 31 0 72 0;
 #X connect 31 1 40 0;
 #X connect 31 2 30 0;
 #X connect 33 0 27 0;
@@ -132,3 +143,8 @@ on Haar's algorithm ( decision tree algorithm );
 #X connect 63 0 68 0;
 #X connect 68 0 31 0;
 #X connect 69 0 13 1;
+#X connect 72 0 26 0;
+#X connect 73 0 31 0;
+#X connect 74 0 73 0;
+#X connect 75 0 69 0;
+#X connect 76 0 31 0;
diff --git a/pdp_opencv_haarcascade.c b/pdp_opencv_haarcascade.c
index 48382a7..1c1edc6 100644
--- a/pdp_opencv_haarcascade.c
+++ b/pdp_opencv_haarcascade.c
@@ -34,6 +34,8 @@
 #include "cv.h"
 #endif
 
+#define MAX_MARKERS 50
+
 const char* default_cascade ="./haarcascade_frontalface_alt.xml";
 
 typedef struct pdp_opencv_haarcascade_struct
@@ -45,6 +47,11 @@ typedef struct pdp_opencv_haarcascade_struct
     t_outlet *x_outlet1;
     t_outlet *x_dataout;
     t_atom rlist[4];
+    int x_xmark[MAX_MARKERS];
+    int x_ymark[MAX_MARKERS];
+    int x_found[MAX_MARKERS];
+    int x_ftolerance;
+
     int x_packet0;
     int x_packet1;
     int x_dropped;
@@ -68,7 +75,30 @@ typedef struct pdp_opencv_haarcascade_struct
     
 } t_pdp_opencv_haarcascade;
 
+static int pdp_opencv_haarcascade_mark(t_pdp_opencv_haarcascade *x, t_floatarg fx, t_floatarg fy )
+{
+  int i;
+  
+    if ( ( fx < 0.0 ) || ( fx > x->x_width ) || ( fy < 0 ) || ( fy > x->x_height ) )
+    {
+       return -1;
+    }
 
+    for ( i=0; i<MAX_MARKERS; i++)
+    {
+       if ( x->x_xmark[i] == -1 )
+       {
+          x->x_xmark[i] = (int)fx;
+          x->x_ymark[i] = (int)fy;
+          x->x_found[i] = x->x_ftolerance;
+          post( "pdp_opencv_haarcascade : inserted point %d (%d,%d)", i, x->x_xmark[i], x->x_ymark[i] );
+          return i;
+       }
+    }
+
+    post( "pdp_opencv_haarcascade : max markers reached" );
+    return -1;
+}
 
 static void pdp_opencv_haarcascade_process_rgb(t_pdp_opencv_haarcascade *x)
 {
@@ -76,7 +106,7 @@ static void pdp_opencv_haarcascade_process_rgb(t_pdp_opencv_haarcascade *x)
     short int *data   = (short int *)pdp_packet_data(x->x_packet0);
     t_pdp     *newheader = pdp_packet_header(x->x_packet1);
     short int *newdata = (short int *)pdp_packet_data(x->x_packet1); 
-    int i;
+    int i, im;
 
     if ((x->x_width != (t_int)header->info.image.width) || 
         (x->x_height != (t_int)header->info.image.height)) 
@@ -148,25 +178,67 @@ static void pdp_opencv_haarcascade_process_rgb(t_pdp_opencv_haarcascade *x)
         else
             outlet_float(x->x_outlet1, 0.0);
 
+        for ( im=0; im<MAX_MARKERS; im++ )
+        {
+           if ( x->x_xmark[im] != -1.0 )
+           {
+             x->x_found[im]--;
+           }
+        }
+
         for( i = 0; i < (faces ? faces->total : 0); i++ )
         {
+          int oi, found;
+
             CvRect* r = (CvRect*)cvGetSeqElem( faces, i );
             CvPoint center;
             int radius;
             center.x = cvRound((r->x + r->width*0.5)*scale);
             center.y = cvRound((r->y + r->height*0.5)*scale);
             radius = cvRound((r->width + r->height)*0.25*scale);
-            cvCircle( x->img, center, radius, colors[i%8], 3, 8, 0 );
+
+            found = 0;
+            oi = -1;
+            for ( im=0; im<MAX_MARKERS; im++ )
+            {
+              // check if the object is already known
+              if ( sqrt( pow(center.x - x->x_xmark[im], 2 ) + pow(center.y - x->x_ymark[im], 2 ) ) <= radius )
+              {
+                 oi=im;
+                 found=1;
+                 x->x_found[im] = x->x_ftolerance;
+                 x->x_xmark[im] = center.x;
+                 x->x_ymark[im] = center.y;
+                 break;
+              }
+            }
+            // new object detected
+            if ( !found )
+            {
+               oi = pdp_opencv_haarcascade_mark(x, center.x, center.y );
+            }
             char tindex[4];
-            sprintf( tindex, "%d", i );
+            sprintf( tindex, "%d", oi );
+            cvCircle( x->img, center, radius, colors[oi%8], 3, 8, 0 );
             cvPutText( x->img, tindex, center, &x->font, CV_RGB(255,255,255));        
 
-            SETFLOAT(&x->rlist[0], i);
+            SETFLOAT(&x->rlist[0], oi);
             SETFLOAT(&x->rlist[1], center.x);
             SETFLOAT(&x->rlist[2], center.y);
             SETFLOAT(&x->rlist[3], radius);
     	    outlet_list( x->x_dataout, 0, 4, x->rlist );
         }
+        // delete lost objects
+        for ( im=0; im<MAX_MARKERS; im++ )
+        {
+           if ( x->x_found[im] < 0 )
+           {
+             x->x_xmark[im] = -1.0;
+             x->x_ymark[im] = -1,0;
+             x->x_found[im] = x->x_ftolerance;
+             post( "deleted point %d", im );
+           }
+        }
     }
 
     cvReleaseImage( &gray );
@@ -198,6 +270,36 @@ static void pdp_opencv_haarcascade_min_neighbors(t_pdp_opencv_haarcascade *x, t_
 	if (f>=1) x->min_neighbors = (int)f;
 }
 
+static void pdp_opencv_haarcascade_ftolerance(t_pdp_opencv_haarcascade *x, t_floatarg f)
+{
+	if (f>=1) x->x_ftolerance = (int)f;
+}
+
+static void pdp_opencv_haarcascade_delete(t_pdp_opencv_haarcascade *x, t_floatarg findex )
+{
+  int i;
+
+    if ( ( findex < 1.0 ) || ( findex > 10 ) )
+    {
+       return;
+    }
+
+    x->x_xmark[(int)findex-1] = -1;
+    x->x_ymark[(int)findex-1] = -1;
+}
+
+static void pdp_opencv_haarcascade_clear(t_pdp_opencv_haarcascade *x )
+{
+  int i;
+
+    for ( i=0; i<MAX_MARKERS; i++)
+    {
+      x->x_xmark[i] = -1;
+      x->x_ymark[i] = -1;
+      x->x_found[i] = x->x_ftolerance;
+    }
+}
+
 static void pdp_opencv_haarcascade_load(t_pdp_opencv_haarcascade *x, t_symbol *filename)
 {
     x->cascade = (CvHaarClassifierCascade*)cvLoad( filename->s_name, 0, 0, 0 );
@@ -221,13 +323,6 @@ static void pdp_opencv_haarcascade_process(t_pdp_opencv_haarcascade *x)
 {
    int encoding;
    t_pdp *header = 0;
-   char *parname;
-   unsigned pi;
-   int partype;
-   float pardefault;
-   t_atom plist[2];
-   t_atom tlist[2];
-   t_atom vlist[2];
 
    /* check if image data packets are compatible */
    if ( (header = pdp_packet_header(x->x_packet0))
@@ -282,7 +377,7 @@ t_class *pdp_opencv_haarcascade_class;
 
 void *pdp_opencv_haarcascade_new(t_floatarg f)
 {
-    int i;
+  int i;
 
     t_pdp_opencv_haarcascade *x = (t_pdp_opencv_haarcascade *)pd_new(pdp_opencv_haarcascade_class);
 
@@ -304,6 +399,13 @@ void *pdp_opencv_haarcascade_new(t_floatarg f)
     x->min_neighbors = 2;
     x->mode = 0;
     x->min_size = 30;
+    x->x_ftolerance = 5;
+
+    for ( i=0; i<MAX_MARKERS; i++ )
+    {
+      x->x_xmark[i] = -1;
+      x->x_ymark[i] = -1;
+    }
 
     x->cascade = (CvHaarClassifierCascade*)cvLoad( default_cascade, 0, 0, 0 );
     if( !x->cascade )
@@ -341,8 +443,8 @@ void pdp_opencv_haarcascade_setup(void)
     class_addmethod(pdp_opencv_haarcascade_class, (t_method)pdp_opencv_haarcascade_min_neighbors, gensym("min_neighbors"),  A_FLOAT, A_NULL );   
     class_addmethod(pdp_opencv_haarcascade_class, (t_method)pdp_opencv_haarcascade_mode, gensym("mode"),  A_FLOAT, A_NULL );   
     class_addmethod(pdp_opencv_haarcascade_class, (t_method)pdp_opencv_haarcascade_min_size, gensym("min_size"),  A_FLOAT, A_NULL );   
-
-
+    class_addmethod(pdp_opencv_haarcascade_class, (t_method)pdp_opencv_haarcascade_ftolerance, gensym("ftolerance"), A_FLOAT, A_NULL );
+    class_addmethod(pdp_opencv_haarcascade_class, (t_method)pdp_opencv_haarcascade_clear, gensym("clear"), A_NULL );
 }
 
 #ifdef __cplusplus