Remove Canny, Rewrote Lane Detection, Changed IPM

README.md

@@ -1 +1,2 @@
 EuroTruckAutopilot
+https://www.youtube.com/watch?v=_s00JKtwZUc

hough_lines.py

@@ -1,3 +1,5 @@
+import datetime
+
 import cv2
 import numpy as np
 
@@ -6,44 +8,55 @@
 
 
 def hough_lines(warped, original):
-    linesP = cv2.HoughLinesP(warped, 1, np.pi / 180, 50, None, 25, 150)
-
-    if linesP is not None:
-        for i in range(0, len(linesP)):
-            #print("# of lines:", len(linesP))
-
-            l = linesP[i][0]
-            x1 = l[0]
-            y1 = l[1]
-            x2 = l[2]
-            y2 = l[3]
-
-            slope = (y2 - y1) / (x2 - x1)
-
-            if slope < 0:
-                for x1, y1, x2, y2 in linesP[i]:
-                    if (np.isnan(x1) == False) and (np.isnan(y1) == False) and (np.isnan(x2) == False) and (
-                            np.isnan(y2) == False):
-                        if ((x2 - x1) != 0) and ((y2 - y1) != 0):
-                            left_avg_slope = (y2 - y1) / (x2 - x1)
-                            left_avg_intercept = y1 - left_avg_slope * x1
-                            left_startx = int((starty - left_avg_intercept) / left_avg_slope)
-                            left_endx = int((endy - left_avg_intercept) / left_avg_slope)
-                            cv2.line(original, (left_startx, starty), (left_endx, endy), (0, 0, 255), 3)
-
-            if slope > 0:
-                for x1, y1, x2, y2 in linesP[i]:
-                    if (np.isnan(x1) == False) and (np.isnan(y1) == False) and (np.isnan(x2) == False) and (
-                            np.isnan(y2) == False):
-                        if ((x2 - x1) != 0) and ((y2 - y1) != 0):
-                            right_avg_slope = (y2 - y1) / (x2 - x1)
-                            right_avg_intercept = y1 - right_avg_slope * x1
-                            right_startx = int((starty - right_avg_intercept) / right_avg_slope)
-                            right_endx = int((endy - right_avg_intercept) / right_avg_slope)
-                            cv2.line(original, (right_startx, starty), (right_endx, endy), (255, 0, 0), 3)
-
-            cv2.namedWindow("Test2", cv2.WINDOW_NORMAL)
-            cv2.resizeWindow("Test2", 500, 500)
-            cv2.moveWindow("Test2", -940, 0)
-            cv2.imshow("Test2", original)
-            cv2.waitKey(1)
+    lines_p = cv2.HoughLinesP(warped, 1, np.pi / 180, 50, None, 50, 300)
+    if lines_p is not None:
+        length = len(lines_p)
+        extrapolated_lines = []
+        start_time = datetime.datetime.now().time().microsecond
+        for i in range(0, length):
+            for j in range(0, length):
+                if i != j:
+                    line_1 = lines_p[i][0]
+                    line_2 = lines_p[j][0]
+
+                    if False not in (line_1.all(), line_2.all()):
+                        line_1_x1 = line_1[0]
+                        line_1_y1 = line_1[1]
+                        line_1_x2 = line_1[2]
+                        line_1_y2 = line_1[3]
+
+                        line_2_x1 = line_2[0]
+                        line_2_y1 = line_2[1]
+                        line_2_x2 = line_2[2]
+                        line_2_y2 = line_2[3]
+
+                        line_x_distance = np.sqrt(np.square(line_2_x1 - line_1_x1) + np.square(line_2_x2 - line_1_x2))
+                        line_y_length = np.abs((line_2_y1 + line_1_y1) / 2 - (line_2_y2 + line_1_y2) / 2)
+                        distance = (np.sqrt(np.square(line_2_x1 - line_1_x1) + np.square(line_2_y1 - line_1_y1)) +
+                                    np.sqrt(np.square(line_2_x2 - line_1_x2) + np.square(line_2_y2 - line_1_y2))) / 2
+
+                        if line_x_distance < 20 and line_y_length > 40 and 10 < distance <= 100:
+                            line_3_x1 = (line_2_x1 + line_1_x1) / 2
+                            if line_3_x1 not in extrapolated_lines:
+                                line_3_y1 = (line_2_y1 + line_1_y1) / 2
+                                line_3_x2 = (line_2_x2 + line_1_x2) / 2
+                                line_3_y2 = (line_2_y2 + line_1_y2) / 2
+
+                                if 0 not in (
+                                        line_3_x1, line_3_x2, line_3_y1, line_3_y2,
+                                        np.abs(line_3_x1) - np.abs(line_3_x2),
+                                        np.abs(line_3_y1) - np.abs(line_3_y2)):
+                                    line_3_avg_slope = (line_3_y2 - line_3_y1) / (line_3_x2 - line_3_x1)
+                                    line_3_avg_intercept = line_3_y1 - line_3_avg_slope * line_3_x1
+                                    line_3_startx = int((starty - line_3_avg_intercept) / line_3_avg_slope)
+                                    right_endx = int((endy - line_3_avg_intercept) / line_3_avg_slope)
+                                    extrapolated_lines.append(line_3_x1)
+                                    cv2.line(original, (line_3_startx, starty), (right_endx, endy), (0, 0, 255), 3)
+
+        print("# of lines:", length, "found:", len(extrapolated_lines),
+              datetime.datetime.now().time().microsecond - start_time)
+        cv2.namedWindow("Test2", cv2.WINDOW_NORMAL)
+        cv2.resizeWindow("Test2", 450, 500)
+        cv2.moveWindow("Test2", -500, 300)
+        cv2.imshow("Test2", original)
+        cv2.waitKey(1)

image_process.py

@@ -1,34 +1,37 @@
 import cv2
 import numpy as np
+
 import hough_lines
 
+
 def pre_process(warped, original):
     blur = cv2.GaussianBlur(warped, (15, 15), 0)
     gray = cv2.cvtColor(blur, cv2.COLOR_BGR2GRAY)
     adaptive_thresh = cv2.adaptiveThreshold(gray, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY, 21, -3)
     merge = cv2.morphologyEx(adaptive_thresh, cv2.MORPH_CLOSE, np.ones((9, 9), np.uint8))
 
-    median = np.median(merge)
-    lower = int(max(0, (1.0 - 0.1) * median))
-    upper = int(min(255, (1.0 + 0.1) * median))
-    canny = cv2.Canny(merge, lower, upper)
+    # median = np.median(merge)
+    # lower = int(max(0, (1.0 - 0.33) * median))
+    # upper = int(min(255, (1.0 + 0.33) * median))
+    # canny = cv2.Canny(merge, lower, upper)
+    # merge = cv2.morphologyEx(canny, cv2.MORPH_CLOSE, np.ones((21, 21), np.uint8))
 
     cv2.namedWindow("adaptiveThreshold", cv2.WINDOW_NORMAL)
     cv2.resizeWindow("adaptiveThreshold", 450, 500)
     cv2.moveWindow("adaptiveThreshold", -1380, 300)
     cv2.imshow("adaptiveThreshold", adaptive_thresh)
     cv2.waitKey(1)
 
-    cv2.namedWindow("canny", cv2.WINDOW_NORMAL)
-    cv2.resizeWindow("canny", 450, 500)
-    cv2.moveWindow("canny", -540, 300)
-    cv2.imshow("canny", canny)
-    cv2.waitKey(1)
+    # cv2.namedWindow("canny", cv2.WINDOW_NORMAL)
+    # cv2.resizeWindow("canny", 450, 500)
+    # cv2.moveWindow("canny", -540, 300)
+    # cv2.imshow("canny", canny)
+    # cv2.waitKey(1)
 
     cv2.namedWindow("merge", cv2.WINDOW_NORMAL)
     cv2.resizeWindow("merge", 450, 500)
     cv2.moveWindow("merge", -940, 300)
     cv2.imshow("merge", merge)
     cv2.waitKey(1)
 
-    #hough_lines.hough_lines(canny, original)
+    hough_lines.hough_lines(merge, original)

perspective_transform.py

@@ -1,11 +1,12 @@
 import cv2
-import numpy as np
 import keyboard
+import numpy as np
+
 import image_process
 
-a = 920
-b = -160
-c = 70
+a = 20
+b = 40
+c = 60
 d = -70
 
 
@@ -53,7 +54,6 @@ def perspective_transform(window):
         d -= 10
         print("d:", d)
 
-
     src = np.float32(
         [[(-a), height - b],
          [width + a, height - b],