1 00:00:01,060 --> 00:00:06,460 In the previous cooling session, we implemented the step one of the localized state, which was extracting 2 00:00:06,460 --> 00:00:07,330 the background model. 3 00:00:07,840 --> 00:00:13,060 Now that we have extracted the bedroom model, we can move on to detecting our car and then extracting 4 00:00:13,060 --> 00:00:15,490 its location using background subtraction. 5 00:00:16,000 --> 00:00:18,730 For this, we move inside the localized board function. 6 00:00:19,120 --> 00:00:24,790 But after we have extracted the background, we implement the step to where it is foreground extraction. 7 00:00:26,290 --> 00:00:31,960 This is for detecting our car and this will be implemented using absolute difference. 8 00:00:31,960 --> 00:00:33,250 The Function of open series. 9 00:00:36,150 --> 00:00:41,250 Now what this function does is simply find the difference between two images and then fix their absolute. 10 00:00:41,670 --> 00:00:49,500 So for the source one, I'm going to be providing the current frame and for the source to our background 11 00:00:49,500 --> 00:00:49,800 world. 12 00:00:52,830 --> 00:00:55,500 This gives us change between both frames. 13 00:00:56,400 --> 00:01:02,130 So we need more times change now because the current Amazon BGR space. 14 00:01:02,340 --> 00:01:07,170 So the change will also be in BGR can space, but we're required in grayscale. 15 00:01:07,290 --> 00:01:11,620 So we will be using the open see function of convert color to convert it into grayscale. 16 00:01:11,640 --> 00:01:17,100 So once we have converted into grayscale, we now extract the mask of all the change objects in the 17 00:01:17,100 --> 00:01:17,460 frame. 18 00:01:17,790 --> 00:01:20,850 So that can be done by using function or thresholding. 19 00:01:21,510 --> 00:01:25,590 So severe to the threshold takes an input argument of sorts. 20 00:01:25,740 --> 00:01:29,100 That is our change group for the thrush. 21 00:01:29,460 --> 00:01:29,790 I'm going. 22 00:01:29,790 --> 00:01:35,820 I'm providing a measly value of 15 just to only ignore the values that are too small to see. 23 00:01:36,540 --> 00:01:43,080 And for the Maxwell the Maxwell you will be providing to 45 because we are looking to perform binary 24 00:01:43,080 --> 00:01:49,620 thresholding to get the mosque to change and type is binary thresh or thresh binary. 25 00:01:50,550 --> 00:01:55,790 And because thresholding provides two outputs and we are only interested in the second output when we 26 00:01:55,800 --> 00:01:58,410 are writing here one to get the second output. 27 00:01:58,980 --> 00:02:04,170 So the placeholder is change mask, which was our output. 28 00:02:04,560 --> 00:02:10,980 Now this change models not only contains the models of object that is the core, but also a lot of random 29 00:02:10,980 --> 00:02:11,340 noise. 30 00:02:11,940 --> 00:02:16,200 So because this noise is random, it will be in very small size. 31 00:02:16,470 --> 00:02:22,050 So we can leverage this information and only take the largest object that is our code and remove all 32 00:02:22,050 --> 00:02:22,500 the noise. 33 00:02:22,740 --> 00:02:26,580 And that can be done by using a function that is present inside this file. 34 00:02:27,180 --> 00:02:30,210 So we import that function that is return largest object. 35 00:02:31,200 --> 00:02:32,220 After importing it. 36 00:02:32,460 --> 00:02:38,460 We simply call that function of return largest object also in the chain models as input. 37 00:02:39,390 --> 00:02:41,010 And this gives me the two outputs. 38 00:02:41,310 --> 00:02:43,260 That is the largest object mask. 39 00:02:43,800 --> 00:02:44,790 That is the car mask. 40 00:02:46,290 --> 00:02:49,050 And the control is control. 41 00:02:50,310 --> 00:02:55,340 Now, once we have the mask and the control of the largest object, we can now extract this location 42 00:02:55,350 --> 00:02:58,380 or should as the relative location of the car to the myth. 43 00:02:59,010 --> 00:03:03,210 To do that, we'll be using a function that is good car location. 44 00:03:04,260 --> 00:03:10,020 Now get car location will take two inputs that it is the car control and the car mask. 45 00:03:11,310 --> 00:03:14,520 So we haven't defined this function yet, so we need to define it first. 46 00:03:14,850 --> 00:03:16,020 Let's define that function. 47 00:03:17,350 --> 00:03:19,300 So that is great collocation. 48 00:03:19,870 --> 00:03:21,370 The first is the car control. 49 00:03:22,660 --> 00:03:27,550 And second, or should I say the first is the self argument and then the car control. 50 00:03:27,550 --> 00:03:29,800 And the third argument is the car must. 51 00:03:32,200 --> 00:03:36,790 Now for the definition of this particular function, the very first thing that you need to do to get 52 00:03:36,790 --> 00:03:40,510 the location of any object, we need to find the centroid of that object. 53 00:03:41,170 --> 00:03:44,680 So we have to define the function, define centroid of object. 54 00:03:45,040 --> 00:03:48,940 So let's define the function that gets the centroid of any contour. 55 00:03:49,090 --> 00:03:50,890 So that is named as the centroid. 56 00:03:51,250 --> 00:03:53,770 What it does is simply takes a control of an object. 57 00:03:54,040 --> 00:04:00,520 Finest moments devise the first moment with the zeroth moment to get the centroid in the form of a tuple. 58 00:04:01,000 --> 00:04:08,590 So let's call this particular function by writing so that centroid mistakes in the input argument of 59 00:04:08,590 --> 00:04:11,710 the card counter and gives us. 60 00:04:12,820 --> 00:04:18,370 The centre of our board, so we name it as the center. 61 00:04:19,540 --> 00:04:23,920 Now that we have the ball center in the final two ball, this is the location of the car with respect 62 00:04:23,920 --> 00:04:24,410 to frame. 63 00:04:24,820 --> 00:04:28,060 But we require the relative location of the car with the means. 64 00:04:28,600 --> 00:04:34,450 So to get to that point, we need to perform the same transformation that we did to the myth that includes 65 00:04:34,450 --> 00:04:41,470 the translation of origin from frame two to the maze, and also the rotation to get the maze entry point 66 00:04:41,470 --> 00:04:41,890 on top. 67 00:04:42,430 --> 00:04:44,650 So to do that, we can't use this double. 68 00:04:44,650 --> 00:04:48,640 We did binary because the transformations need to be done on an entry. 69 00:04:48,880 --> 00:04:55,660 So we need to convert this tuple into an array and also we need to flip this tuple from and why x format 70 00:04:55,660 --> 00:05:01,260 to x y form, which is the formatting which which is used to perform all the transformations. 71 00:05:01,270 --> 00:05:03,280 So let's do both these things together. 72 00:05:03,580 --> 00:05:10,900 So right here we have simply converted the ball center that is a tuple to an array and also flip from 73 00:05:10,900 --> 00:05:13,330 y experiment to an x y format. 74 00:05:13,840 --> 00:05:18,070 We're taking the second value in the first and the first value the second. 75 00:05:18,670 --> 00:05:20,030 This gives us the ball center. 76 00:05:20,030 --> 00:05:25,170 Right now we have it in the format that we can now perform the transformations in the same way that 77 00:05:25,210 --> 00:05:27,400 form the transformations for our maze. 78 00:05:27,880 --> 00:05:30,940 So the first transformation is the translation of the origin. 79 00:05:31,390 --> 00:05:37,320 So that can be done by simply creating a placeholder at this box. 80 00:05:37,330 --> 00:05:45,430 Enter translated and that can be written as non point or zeros like the temporary. 81 00:05:48,110 --> 00:05:49,890 So we now have the placeholder. 82 00:05:49,910 --> 00:05:51,710 We simply need to do the translation. 83 00:05:52,190 --> 00:05:55,280 So that is simply written as more center translated. 84 00:05:55,760 --> 00:05:57,380 So we translate is force value. 85 00:05:57,590 --> 00:05:58,480 That is the x. 86 00:05:59,970 --> 00:06:01,140 By taking the ball central. 87 00:06:10,540 --> 00:06:19,090 While central areas first value and simply subtracting this with the origin origin column. 88 00:06:19,480 --> 00:06:25,990 So subtraction the more center column with the origin column simply gives us the translated more center 89 00:06:25,990 --> 00:06:28,960 column in the same way we'll be translating the rule. 90 00:06:28,960 --> 00:06:35,200 And once we closer to the row, we can now move on to the next step that is performing the rotation 91 00:06:35,380 --> 00:06:37,210 on our translated board center. 92 00:06:39,790 --> 00:06:46,270 So that we perform in the same way that we did with the translation that is using the previously saved 93 00:06:46,360 --> 00:06:51,030 or updated parameters that we did on that, we did transformations on our Ms.. 94 00:06:51,520 --> 00:06:52,870 So that is for rotation. 95 00:06:52,870 --> 00:06:54,610 We have the rotation metrics. 96 00:06:55,090 --> 00:07:00,930 So we use that rotation metrics, use it to transform our ball center translator or rotate our bottom, 97 00:07:01,030 --> 00:07:04,180 translate it in the same way that we rotated our means. 98 00:07:05,050 --> 00:07:12,190 And then we transpose this and then transpose again to get this in the form of a nice little point. 99 00:07:12,610 --> 00:07:17,050 That is what location on Miss. 100 00:07:17,680 --> 00:07:21,160 So we now have the board relative location on the miss. 101 00:07:22,090 --> 00:07:22,420 So. 102 00:07:23,470 --> 00:07:24,910 They should completed right here. 103 00:07:25,150 --> 00:07:31,180 But because of rotation, there might come a point that is negative in its value, but location can 104 00:07:31,180 --> 00:07:31,990 never be negative. 105 00:07:32,200 --> 00:07:34,840 This is because the origin is shifted in the negative direction. 106 00:07:35,200 --> 00:07:36,860 So we want a positive value. 107 00:07:36,880 --> 00:07:39,190 So we need to translate this back. 108 00:07:39,190 --> 00:07:40,960 Or if it goes to negative. 109 00:07:42,360 --> 00:07:42,750 So. 110 00:07:43,970 --> 00:07:51,440 By simply adding the column rotation column to our core rules and adding the rules to our column, we 111 00:07:51,440 --> 00:07:54,440 can simply translate the origin if it gets to negative values. 112 00:07:55,040 --> 00:08:00,530 Now, once you've done this, simply need to save it inside the placeholder of the block location on 113 00:08:00,530 --> 00:08:00,920 the Miss. 114 00:08:01,130 --> 00:08:06,800 So that placeholder name is self-taught location GA and this stores the location of the core or relative 115 00:08:06,800 --> 00:08:08,390 location of the car on the miss. 116 00:08:09,110 --> 00:08:12,590 So we need to name this placeholder inside init function. 117 00:08:13,880 --> 00:08:14,690 So right here. 118 00:08:14,840 --> 00:08:16,160 So location, car. 119 00:08:17,000 --> 00:08:18,380 And it's like this to zero. 120 00:08:18,710 --> 00:08:21,920 So this will store the location of the car according to the myth. 121 00:08:22,400 --> 00:08:26,630 So now we know that we have completed the function of getting our location. 122 00:08:27,710 --> 00:08:31,640 We have called it right here in the localize but the main function of our localized state. 123 00:08:31,910 --> 00:08:37,580 We can move on to simply join the detected car so that we can see the user can see where the car is 124 00:08:37,580 --> 00:08:38,370 at any moment. 125 00:08:38,390 --> 00:08:44,510 So the way we do this is simply find the enclosing circle for the car control, draw that circle, and 126 00:08:44,510 --> 00:08:47,330 then simply display in the phone for a nice looking red color. 127 00:08:48,140 --> 00:08:51,410 Once we have done this, we display our detected car. 128 00:08:52,250 --> 00:08:58,220 So the localized car is shown on the frame display and also the foreground is being displayed clearly. 129 00:08:59,030 --> 00:09:03,710 So now we run this particular file around this project to see whether we have correctly implemented 130 00:09:03,920 --> 00:09:08,660 the state or the localized state for our problem solving. 131 00:09:08,930 --> 00:09:09,260 What? 132 00:09:11,280 --> 00:09:12,450 So we go inside the terminal. 133 00:09:14,140 --> 00:09:14,890 Will the project. 134 00:09:16,830 --> 00:09:18,010 Launch a simulation. 135 00:09:20,050 --> 00:09:24,640 So once the simulation starts up, we run our maze solver file. 136 00:09:25,700 --> 00:09:30,860 And as you can see, the male swallow part has run and it has correctly extracted a full step. 137 00:09:30,890 --> 00:09:32,030 It is a cracking starter. 138 00:09:32,270 --> 00:09:39,530 The males in the phone for me as entry on top and then perform just the next step that is stuck in the 139 00:09:39,530 --> 00:09:40,040 foreground. 140 00:09:40,130 --> 00:09:45,500 As you can see, the car has been correctly extracted from all of the noise and noise have been removed 141 00:09:45,800 --> 00:09:47,240 and only the car has been removed. 142 00:09:48,470 --> 00:09:49,850 And this is the mass of our car. 143 00:09:50,390 --> 00:09:55,280 So if you just look at the frame display, you can see that the car has been correctly localize right 144 00:09:55,280 --> 00:09:58,730 at the entry, the red circle drone on top of it. 145 00:09:59,180 --> 00:10:04,460 So we have found the location of the car relative to the means and this means we have correctly implemented 146 00:10:04,700 --> 00:10:10,730 the localised state of our assault and what so we can move on to the next step that is attempting the 147 00:10:10,730 --> 00:10:11,450 second stage. 148 00:10:11,480 --> 00:10:13,730 That is a mapping for our missile. 149 00:10:13,730 --> 00:10:15,050 Not till then. 150 00:10:15,290 --> 00:10:15,680 Thank you. 151 00:10:15,680 --> 00:10:16,070 And by.