{ "cells": [ { "cell_type": "markdown", "metadata": { "_cell_guid": "7777d9a4-3741-606b-af50-12cadc4c5717" }, "source": [ "Using pattern matching to find the red dotted sea lions." ] }, { "cell_type": "markdown", "metadata": { "_cell_guid": "9b8cbb31-b7a2-929c-3ff5-e0bd808f28f0" }, "source": [ "This version works a lot better by filtering out the red pixels first, then finding circles in the resulting image. " ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "1140cf35-22c3-43fa-8de9-f5c4fcd7b121" }, "outputs": [], "source": [ "import numpy as np\n", "import cv2\n", "import matplotlib.pyplot as plt\n", "\n", "img = cv2.cvtColor(cv2.imread(\"../input/TrainDotted/5.jpg\"), cv2.COLOR_BGR2RGB)\n", "plt.imshow(img, cmap='gray')" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "50e7b13a-af58-9e2e-edab-cec89fa5875e" }, "outputs": [], "source": [ "# Mask everything but the red dots.\n", "cmsk = cv2.inRange(img, np.array([160, 0, 0]), np.array([255, 50, 50])) # Get the red -ish stuff.\n", "# Find the circles in the masked image.\n", "circles = cv2.HoughCircles(cmsk,cv2.HOUGH_GRADIENT,1,50, param1=40,param2=1,minRadius=0,maxRadius=25)" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "5a725813-303e-aeb6-44fb-1656ed362b2e" }, "outputs": [], "source": [ "#Draw rectangles around the dots. Print the total.\n", "'''\n", "if circles is not None:\n", " circles = np.uint16(np.around(circles))\n", " print('%d circles found.' % (len(circles[0]) ))\n", " for i in circles[0,:]:\n", " cv2.rectangle(img, (i[0] - 50, i[1] - 50), (i[0] + 50, i[1] + 50), (255, 0, 0), 3)\n", "\n", "plt.imshow(img, cmap = 'gray', interpolation = 'bicubic')\n", "'''" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "a3f32d23-ed8c-b755-57fe-947eeb1ea4f4" }, "outputs": [], "source": [ "#Extract one sealion\n", "i=0 #Sealion number\n", "print(circles[0,i,:])\n", "imgCrop= img[int(circles[0,i,1]-50):int(circles[0,i,1]+50),int(circles[0,i,0]-50):int(circles[0,i,0]+50)]\n", "plt.imshow(imgCrop, cmap = 'gray', interpolation = 'bicubic')\n", "\n", "#Convert to grayscal and normalize\n", "imgCrop=cv2.cvtColor(imgCrop, cv2.COLOR_RGB2GRAY)\n", "cv2.normalize(imgCrop, imgCrop, 0, 255, cv2.NORM_MINMAX)\n", "plt.imshow(imgCrop, cmap = 'gray', interpolation = 'bicubic')" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "d0eb3fb8-b388-b97c-9e3d-d773375d08af" }, "outputs": [], "source": [ "#Apply a canny filter to extract the edges\n", "edges = cv2.Canny(imgCrop,50,50)\n", "plt.imshow(edges,cmap = 'gray')" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "e4883437-476a-c335-fe0a-bcb32c1e08c5" }, "outputs": [], "source": [ "kernel = np.ones((5,5),np.uint8)\n", "#dilation = cv2.dilate(edges,kernel,iterations = 1)\n", "dilation = cv2.morphologyEx(edges, cv2.MORPH_CLOSE, kernel)\n", "plt.imshow(dilation,cmap = 'gray')" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "915b04db-6b49-284c-5a25-44e77a910618" }, "outputs": [], "source": [ "#Fill function\n", "def fillEdgeImage(img):\n", " edgesNeg = img.copy();\n", " # Mask used to flood filling.\n", " # Notice the size needs to be 2 pixels than the image.\n", " h, w = img.shape[:2]\n", " mask = np.zeros((h+2, w+2), np.uint8)\n", " \n", " # Floodfill from point (0, 0)\n", " cv2.floodFill(edgesNeg, mask, (0,0), 255);\n", " cv2.bitwise_not(edgesNeg, edgesNeg);\n", " filledImg = (edgesNeg | img);\n", " return filledImg" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "6892e946-3fd8-b0c6-5363-f82eb1801e6c" }, "outputs": [], "source": [ "filledImg = fillEdgeImage(dilation)\n", "plt.imshow(filledImg,cmap = 'gray')" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "6339018a-e8c0-1694-1f5a-ff5ff0491505" }, "outputs": [], "source": [ "\n", "# Setup SimpleBlobDetector parameters.\n", "params = cv2.SimpleBlobDetector_Params()\n", "\n", "# To extract white blobs\n", "params.blobColor = 255;\n", "# Change thresholds\n", "params.minThreshold = 10\n", "params.maxThreshold = 200\n", "# Filter by Area.\n", "params.filterByArea = True\n", "params.minArea = 0\n", "# Filter by Circularity\n", "params.filterByCircularity = False\n", "params.minCircularity = 0.1\n", "# Filter by Convexity\n", "params.filterByConvexity = False\n", "params.minConvexity = 0.87\n", "# Filter by Inertia\n", "params.filterByInertia = False\n", "params.minInertiaRatio = 0.01\n", "# Create a detector with the parameters\n", "detector = cv2.SimpleBlobDetector_create(params)\n", "\n", "# Detect blobs.\n", "keypoints = detector.detect(filledImg)\n", "print(keypoints)\n", "# Draw detected blobs as red circles.\n", "# cv2.DRAW_MATCHES_FLAGS_DRAW_RICH_KEYPOINTS ensures\n", "# the size of the circle corresponds to the size of blob\n", "\n", "point1=keypoints[0].pt\n", "print(point1)\n", "print(keypoints[0].size)\n", "'''\n", "# Draw detected blobs as red circles.\n", "# cv2.DRAW_MATCHES_FLAGS_DRAW_RICH_KEYPOINTS ensures the size of the circle corresponds to the size of blob\n", "im_with_keypoints = cv2.drawKeypoints(im, keypoints, np.array([]), (0,0,255), cv2.DRAW_MATCHES_FLAGS_DRAW_RICH_KEYPOINTS)\n", "\n", "# Show keypoints\n", "cv2.imshow(\"Keypoints\", im_with_keypoints)\n", "'''" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "f79e8b36-880f-9437-4877-587f12aaa1b5" }, "outputs": [], "source": [ "#Find the blob that is closer to the center of the image\n", "h, w = filledImg.shape\n", "count=0;\n", "euclDistPre=h**2+w**2;\n", "for i in keypoints:\n", " hdiff=h/2-i.pt[0]\n", " wdiff=w/2-i.pt[1]\n", " euclDist = hdiff**2 + wdiff**2\n", " if euclDist