RealSense D435 का उपयोग कर वस्तुओं के लिए दूरी को मापने

आवश्यकताएँ

pip install pyrealsense2

ऑब्जेक्ट डिटेक्शन

1. github.com/opencv/opencv/blob/master/samples/dnn/tf_text_graph_ssd.py
2. github.com/opencv/opencv/blob/master/samples/dnn/tf_text_graph_common.py

python tf_text_graph_ssd.py --input /path/to/model.pb --config /path/to/example.config --output /path/to/graph.pbtxt

विकास

import pyrealsense2 as rs
import numpy as np
import cv2 as cv

def draw_predictions(pred_img, color_img, depth_image):
    pass

pipeline = rs.pipeline() # <-  pipeline      
config = rs.config() # <-      
colorizer = rs.colorizer() # <-      

spatial = rs.spatial_filter()
spatial.set_option(rs.option.holes_fill, 3)

#  
model = cv.dnn.readNetFromTensorflow(
    "models/ssdlite_mobilenet_v2_coco_2018_05_09/frozen_inference_graph.pb", 
    "models/ssdlite_mobilenet_v2_coco_2018_05_09/graph.pbtxt"
)

#  2 : RGB  depth
config.enable_stream(rs.stream.color, 640, 480, rs.format.bgr8, 30)
config.enable_stream(rs.stream.depth, 640, 480, rs.format.z16, 30)
profile = pipeline.start(config)

cv.namedWindow("RealSense object detection", cv.WINDOW_AUTOSIZE)
try:
    while True:
        #     ""  ""
        frames = pipeline.wait_for_frames()
        depth_frame = frames.get_depth_frame()
        color_frame = frames.get_color_frame()
        if not depth_frame or not color_frame:
            continue

        #    numpy-
        depth_image = np.asanyarray(depth_frame.get_data())
        color_image = np.asanyarray(color_frame.get_data())

        #  
        model.setInput(cv.dnn.blobFromImage(color_image, size=(300, 300), swapRB=True, crop=False))
        pred = model.forward()
        draw_predictions(pred, color_image, depth_image)

	#     
        colorized_depth = cv.applyColorMap(cv.convertScaleAbs(depth_image, alpha=0.03), cv.COLORMAP_JET)
        #    
        images = np.hstack((color_image, colorized_depth))
        cv.imshow("RealSense object detection", images)

	#    'ESC'  'q'
        key = cv.waitKey(1) & 0xFF
        if (key == 27) or (key == ord('q')):
            cv.destroyWindow("RealSense object detection")
            break        

finally:
    pipeline.stop()

def draw_predictions(pred_img, color_img, depth_image):
    for detection in pred_img[0,0,:,:]:
        score = float(detection[2])
        #            50%
        if score > 0.5:
            left = detection[3] * color_img.shape[1]
            top = detection[4] * color_img.shape[0]
            right = detection[5] * color_img.shape[1]
            bottom = detection[6] * color_img.shape[0]
            cv.rectangle(color_img, (int(left), int(top)), (int(right), int(bottom)), (210, 230, 23), 2)
            
            #    
            depth = depth_image[int(left):int(right), int(top):int(bottom)].astype(float)
            depth_scale = profile.get_device().first_depth_sensor().get_depth_scale()
            depth = depth * depth_scale
            dist,_,_,_ = cv.mean(depth)
            dist = round(dist, 1)
            cv.putText(color_img, "dist: "+str(dist)+"m", (int(left), int(top)-5), cv.FONT_HERSHEY_PLAIN, 1, (0, 255, 0), 1)

import pyrealsense2 as rs
import numpy as np
import cv2 as cv

def draw_predictions(pred_img, color_img, depth_image): # <-        
    for detection in pred_img[0,0,:,:]:
        score = float(detection[2])
        #            50%
        if score > 0.5:
            left = detection[3] * color_img.shape[1]
            top = detection[4] * color_img.shape[0]
            right = detection[5] * color_img.shape[1]
            bottom = detection[6] * color_img.shape[0]
            cv.rectangle(color_img, (int(left), int(top)), (int(right), int(bottom)), (210, 230, 23), 2)
            
            #    
            depth = depth_image[int(left):int(right), int(top):int(bottom)].astype(float)
            depth_scale = profile.get_device().first_depth_sensor().get_depth_scale()
            depth = depth * depth_scale
            dist,_,_,_ = cv.mean(depth)
            dist = round(dist, 1)
            cv.putText(color_img, "dist: "+str(dist)+"m", (int(left), int(top)-5), cv.FONT_HERSHEY_PLAIN, 1, (0, 255, 0), 1)


pipeline = rs.pipeline() # <-  pipeline      
config = rs.config() # <-      
colorizer = rs.colorizer() # <-      

#  
model = cv.dnn.readNetFromTensorflow(
    "models/ssdlite_mobilenet_v2_coco_2018_05_09/frozen_inference_graph.pb", 
    "models/ssdlite_mobilenet_v2_coco_2018_05_09/graph.pbtxt"
)

#  2 : RGB  depth
config.enable_stream(rs.stream.color, 640, 480, rs.format.bgr8, 30)
config.enable_stream(rs.stream.depth, 640, 480, rs.format.z16, 30)
profile = pipeline.start(config)

cv.namedWindow("RealSense object detection", cv.WINDOW_AUTOSIZE)
try:
    while True:
        #     ""  ""
        frames = pipeline.wait_for_frames()
        depth_frame = frames.get_depth_frame()
        color_frame = frames.get_color_frame()
        if not depth_frame or not color_frame:
            continue

        #    numpy-
        depth_image = np.asanyarray(depth_frame.get_data())
        color_image = np.asanyarray(color_frame.get_data())
        
        #  
        model.setInput(cv.dnn.blobFromImage(color_image, size=(300, 300), swapRB=True, crop=False))
        pred = model.forward()
        draw_predictions(pred, color_image, depth_image)

        #     
        colorized_depth = cv.applyColorMap(cv.convertScaleAbs(depth_image, alpha=0.03), cv.COLORMAP_JET)
        #    
        images = np.hstack((color_image, colorized_depth))
        cv.imshow("RealSense object detection", images)
        
        #    'ESC'  'q'
        key = cv.waitKey(1) & 0xFF
        if (key == 27) or (key == ord('q')):
            cv.destroyWindow("RealSense object detection")
            break

finally:
    pipeline.stop()

निष्कर्ष