五、 3588_kylin_RTSP数据识别

hileev
已于 2024-07-18 09:57:11 修改
阅读量310
点赞数 3
文章标签： kylin 大数据
于 2024-07-18 09:54:13 首次发布
本文链接：https://blog.csdn.net/weixin_40974083/article/details/140514150
版权
#!/usr/bin/env python3
import sys
sys.path.append('/opt/ros/kinetic/lib/python2.7/dist-packages')
sys.path.append('/usr/lib/python2.7/dist-packages')
import rospy
import time
from sensor_msgs.msg import Image
from cv_bridge import CvBridge,CvBridgeError
import numpy as np
import sys

python3 = True if sys.hexversion > 0x03000000 else False
if python3:
    # remove ros python2 library
    ros_cv_path = '/opt/ros/kinetic/lib/python2.7/dist-packages'
    if ros_cv_path in sys.path:
        sys.path.remove(ros_cv_path)

rospy.init_node('listener', anonymous=True)
import cv2 
import os
import time
import copy
from  pathlib import Path
import platform
from rknnlite.api import RKNNLite
#----------------------------------------------------
# decice tree for rk356x/rk3588
DEVICE_COMPATIBLE_NODE = '/proc/device-tree/compatible'
img = []

def get_host():
    # get platform and device type
    system = platform.system()
    machine = platform.machine()
    os_machine = system + '-' + machine
    if os_machine == 'Linux-aarch64':
        try:
            with open(DEVICE_COMPATIBLE_NODE) as f:
                device_compatible_str = f.read()
                if 'rk3588' in device_compatible_str:
                    host = 'RK3588'
                else:
                    host = 'RK356x'
        except IOError:
            print('Read device node {} failed.'.format(DEVICE_COMPATIBLE_NODE))
            exit(-1)
    else:
        host = os_machine
    return host

RK356X_RKNN_MODEL = 'resnet18_for_rk356x.rknn'
RK3588_RKNN_MODEL = '/home/kylin/桌面/newfolder/Yolov5_rknnlite2-main/weights/yolov5s-640-640.rknn'

OBJ_THRESH = 0.25
NMS_THRESH = 0.45
IMG_SIZE = 640

CLASSES = ("person",)

def sigmoid(x):
    return 1 / (1 + np.exp(-x))

def xywh2xyxy(x):
    # Convert [x, y, w, h] to [x1, y1, x2, y2]
    y = np.copy(x)
    y[:, 0] = x[:, 0] - x[:, 2] / 2  # top left x
    y[:, 1] = x[:, 1] - x[:, 3] / 2  # top left y
    y[:, 2] = x[:, 0] + x[:, 2] / 2  # bottom right x
    y[:, 3] = x[:, 1] + x[:, 3] / 2  # bottom right y
    return y

def process(input, mask, anchors):
    anchors = [anchors[i] for i in mask]
    grid_h, grid_w = map(int, input.shape[0:2])

    box_confidence = sigmoid(input[..., 4])
    box_confidence = np.expand_dims(box_confidence, axis=-1)

    box_class_probs = sigmoid(input[..., 5:])

    box_xy = sigmoid(input[..., :2]) * 2 - 0.5

    col = np.tile(np.arange(0, grid_w), grid_w).reshape(-1, grid_w)
    row = np.tile(np.arange(0, grid_h).reshape(-1, 1), grid_h)
    col = col.reshape(grid_h, grid_w, 1, 1).repeat(3, axis=-2)
    row = row.reshape(grid_h, grid_w, 1, 1).repeat(3, axis=-2)
    grid = np.concatenate((col, row), axis=-1)
    box_xy += grid
    box_xy *= int(IMG_SIZE / grid_h)

    box_wh = pow(sigmoid(input[..., 2:4]) * 2, 2)
    box_wh = box_wh * anchors

    box = np.concatenate((box_xy, box_wh), axis=-1)

    return box, box_confidence, box_class_probs

def filter_boxes(boxes, box_confidences, box_class_probs):
    """Filter boxes with box threshold. It's a bit different with origin yolov5 post process!

    # Arguments
        boxes: ndarray, boxes of objects.
        box_confidences: ndarray, confidences of objects.
        box_class_probs: ndarray, class_probs of objects.

    # Returns
        boxes: ndarray, filtered boxes.
        classes: ndarray, classes for boxes.
        scores: ndarray, scores for boxes.
    """
    boxes = boxes.reshape(-1, 4)
    box_confidences = box_confidences.reshape(-1)
    box_class_probs = box_class_probs.reshape(-1, box_class_probs.shape[-1])

    _box_pos = np.where(box_confidences >= OBJ_THRESH)
    boxes = boxes[_box_pos]
    box_confidences = box_confidences[_box_pos]
    box_class_probs = box_class_probs[_box_pos]

    class_max_score = np.max(box_class_probs, axis=-1)
    classes = np.argmax(box_class_probs, axis=-1)
    _class_pos = np.where(class_max_score >= OBJ_THRESH)

    boxes = boxes[_class_pos]
    classes = classes[_class_pos]
    scores = (class_max_score * box_confidences)[_class_pos]

    return boxes, classes, scores

def nms_boxes(boxes, scores):
    """Suppress non-maximal boxes.

    # Arguments
        boxes: ndarray, boxes of objects.
        scores: ndarray, scores of objects.

    # Returns
        keep: ndarray, index of effective boxes.
    """
    x = boxes[:, 0]
    y = boxes[:, 1]
    w = boxes[:, 2] - boxes[:, 0]
    h = boxes[:, 3] - boxes[:, 1]

    areas = w * h
    order = scores.argsort()[::-1]

    keep = []
    while order.size > 0:
        i = order[0]
        keep.append(i)

        xx1 = np.maximum(x[i], x[order[1:]])
        yy1 = np.maximum(y[i], y[order[1:]])
        xx2 = np.minimum(x[i] + w[i], x[order[1:]] + w[order[1:]])
        yy2 = np.minimum(y[i] + h[i], y[order[1:]] + h[order[1:]])

        w1 = np.maximum(0.0, xx2 - xx1 + 0.00001)
        h1 = np.maximum(0.0, yy2 - yy1 + 0.00001)
        inter = w1 * h1

        ovr = inter / (areas[i] + areas[order[1:]] - inter)
        inds = np.where(ovr <= NMS_THRESH)[0]
        order = order[inds + 1]
    keep = np.array(keep)
    return keep

def yolov5_post_process(input_data):
    masks = [[0, 1, 2], [3, 4, 5], [6, 7, 8]]
    anchors = [[10, 13], [16, 30], [33, 23], [30, 61], [62, 45],
               [59, 119], [116, 90], [156, 198], [373, 326]]

    boxes, classes, scores = [], [], []
    for input, mask in zip(input_data, masks):
        b1, c1, s1 = process(input, mask, anchors)
        b1, c1, s1 = filter_boxes(b1, c1, s1)
        boxes.append(b1)
        classes.append(c1)
        scores.append(s1)

    boxes = np.concatenate(boxes)
    boxes = xywh2xyxy(boxes)
    classes = np.concatenate(classes)
    scores = np.concatenate(scores)

    nboxes, nclasses, nscores = [], [], []
    for c in set(classes):
        if c != 0:
            continue
        inds = np.where(classes == c)
        b = boxes[inds]
        c = classes[inds]
        s = scores[inds]

        keep = nms_boxes(b, s)

        nboxes.append(b[keep])
        nclasses.append(c[keep])
        nscores.append(s[keep])

    if not nclasses and not nscores:
        return None, None, None

    boxes = np.concatenate(nboxes)
    classes = np.concatenate(nclasses)
    scores = np.concatenate(nscores)

    return boxes, classes, scores

def draw(image, boxes, scores, classes):
    """Draw the boxes on the image.

    # Argument:
        image: original image.
        boxes: ndarray, boxes of objects.
        classes: ndarray, classes of objects.
        scores: ndarray, scores of objects.
        all_classes: all classes name.
    """
    for box, score, cl in zip(boxes, scores, classes):
        top, left, right, bottom = box
        print('class: {}, score: {}'.format(CLASSES[cl], score))
        print('box coordinate left,top,right,down: [{}, {}, {}, {}]'.format(top, left, right, bottom))
        top = int(top)
        left = int(left)
        right = int(right)
        bottom = int(bottom)

        cv2.rectangle(image, (top, left), (right, bottom), (255, 0, 0), 2)
        cv2.putText(image, '{0} {1:.2f}'.format(CLASSES[cl], score),
                    (top, left - 6),
                    cv2.FONT_HERSHEY_SIMPLEX,
                    0.6, (0, 0, 255), 2)

def letterbox(im, new_shape=(640, 640), color=(0, 0, 0)):
    # Resize and pad image while meeting stride-multiple constraints
    shape = im.shape[:2]  # current shape [height, width]
    if isinstance(new_shape, int):
        new_shape = (new_shape, new_shape)

    # Scale ratio (new / old)
    r = min(new_shape[0] / shape[0], new_shape[1] / shape[1])

    # Compute padding
    ratio = r, r  # width, height ratios
    new_unpad = int(round(shape[1] * r)), int(round(shape[0] * r))
    dw, dh = new_shape[1] - new_unpad[0], new_shape[0] - new_unpad[1]  # wh padding

    dw /= 2  # divide padding into 2 sides
    dh /= 2

    if shape[::-1] != new_unpad:  # resize
        im = cv2.resize(im, new_unpad, interpolation=cv2.INTER_LINEAR)
    top, bottom = int(round(dh - 0.1)), int(round(dh + 0.1))
    left, right = int(round(dw - 0.1)), int(round(dw + 0.1))
    im = cv2.copyMakeBorder(im, top, bottom, left, right, cv2.BORDER_CONSTANT, value=color)  # add border
    return im, ratio, (dw, dh)

#---------------------------------------------------
cv_type = dict()
cv_type['mono8'] = (np.uint8  , 1)
cv_type['8UC2']  = (np.uint8  , 2)
cv_type['bgr8']  = (np.uint8  , 3)
cv_type['8UC4']  = (np.uint8  , 4)
cv_type['8SC1']  = (np.int8   , 1)
cv_type['8SC2']  = (np.int8   , 2)
cv_type['8SC3']  = (np.int8   , 3)
cv_type['8SC4']  = (np.int8   , 4)
cv_type['16UC1'] = (np.uint16 , 1)
cv_type['16UC2'] = (np.uint16 , 2)
cv_type['16UC3'] = (np.uint16 , 3)
cv_type['16UC4'] = (np.uint16 , 4)
cv_type['16SC1'] = (np.int16  , 1)
cv_type['16SC2'] = (np.int16  , 2)
cv_type['16SC3'] = (np.int16  , 3)
cv_type['16SC4'] = (np.int16  , 4)
cv_type['32SC1'] = (np.uint32 , 1)
cv_type['32SC2'] = (np.uint32 , 2)
cv_type['32SC3'] = (np.uint32 , 3)
cv_type['32SC4'] = (np.uint32 , 4)
cv_type['32FC1'] = (np.int32  , 1)
cv_type['32FC2'] = (np.int32  , 2)
cv_type['32FC3'] = (np.int32  , 3)
cv_type['32FC4'] = (np.int32  , 4)
cv_type['64FC1'] = (np.float64, 1)
cv_type['64FC2'] = (np.float64, 2)
cv_type['64FC3'] = (np.float64, 3)
cv_type['64FC4'] = (np.float64, 4)
ros_type = dict()
ros_type[('uint8'  , 1)] = ('mono8', 1)
ros_type[('uint8'  , 2)] = ('8UC2' , 2)
ros_type[('uint8'  , 3)] = ('bgr8' , 3)
ros_type[('uint8'  , 4)] = ('8UC4' , 4)
ros_type[('int8'   , 1)] = ('8SC1' , 1)
ros_type[('int8'   , 2)] = ('8SC2' , 2)
ros_type[('int8'   , 3)] = ('8SC3' , 3)
ros_type[('int8'   , 4)] = ('8SC4' , 4)
ros_type[('uint16' , 1)] = ('16UC1', 1)
ros_type[('uint16' , 2)] = ('16UC2', 2)
ros_type[('uint16' , 3)] = ('16UC3', 3)
ros_type[('uint16' , 4)] = ('16UC4', 4)
ros_type[('int16'  , 1)] = ('16SC1', 1)
ros_type[('int16'  , 2)] = ('16SC2', 2)
ros_type[('int16'  , 3)] = ('16SC3', 3)
ros_type[('int16'  , 4)] = ('16SC4', 4)
ros_type[('uint32' , 1)] = ('32SC1', 1)
ros_type[('uint32' , 2)] = ('32SC2', 2)
ros_type[('uint32' , 3)] = ('32SC3', 3)
ros_type[('uint32' , 4)] = ('32SC4', 4)
ros_type[('int32'  , 1)] = ('32FC1', 1)
ros_type[('int32'  , 2)] = ('32FC2', 2)
ros_type[('int32'  , 3)] = ('32FC3', 3)
ros_type[('int32'  , 4)] = ('32FC4', 4)
ros_type[('float64', 1)] = ('64FC1', 1)
ros_type[('float64', 2)] = ('64FC2', 2)
ros_type[('float64', 3)] = ('64FC3', 3)
ros_type[('float64', 4)] = ('64FC4', 4)

def ToRosMsg(img):
    msg = Image()
    msg.header.stamp = rospy.Time.now()
    msg.width        = img.shape[1]
    msg.height       = img.shape[0]
    channel          = 1
    if len(img.shape) is 3:
        channel = img.shape[2]
    msg.encoding = ros_type[(str(img.dtype), channel)][0]
    msg.is_bigendian = 0
    msg.step = msg.width * channel
    msg.data = img.tobytes()
    return msg

def ToCvImg(msg):
    img = np.fromstring(
        msg.data, dtype=cv_type[msg.encoding][0]).reshape(
            msg.height, msg.width, cv_type[msg.encoding][1])
    return img

def callback(imgmsg):
    img = ToCvImg(imgmsg)
    img = cv2.resize(img,(640,416),interpolation=cv2.INTER_NEAREST)
    host_name = get_host()
    if host_name == 'RK356x':
        rknn_model = RK356X_RKNN_MODEL
    elif host_name == 'RK3588':
        rknn_model = RK3588_RKNN_MODEL
    else:
        print("This demo cannot run on the current platform: {}".format(host_name))
        exit(-1)

    rknn_lite = RKNNLite()
    # load RKNN model
    print('--> Load RKNN model')
    ret = rknn_lite.load_rknn(rknn_model)
    if ret != 0:
        print('Load RKNN model failed')
        exit(ret)
    print('done')

    # init runtime environment
    #print('--> Init runtime environment')
    # run on RK356x/RK3588 with Debian OS, do not need specify target.
    if host_name == 'RK3588':
        ret = rknn_lite.init_runtime(core_mask=RKNNLite.NPU_CORE_0)
    else:
        ret = rknn_lite.init_runtime()
    if ret != 0:
        print('Init runtime environment failed')
        exit(ret)
    print('done')

    img, ratio, (dw, dh) = letterbox(img, new_shape=(IMG_SIZE, IMG_SIZE))
    img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
    print('--> Running model')
    outputs = rknn_lite.inference(inputs=[img])
    input0_data = outputs[0]
    input1_data = outputs[1]
    input2_data = outputs[2]
    input0_data = input0_data.reshape([3, -1] + list(input0_data.shape[-2:]))
    input1_data = input1_data.reshape([3, -1] + list(input1_data.shape[-2:]))
    input2_data = input2_data.reshape([3, -1] + list(input2_data.shape[-2:]))
    input_data = list()
    input_data.append(np.transpose(input0_data, (2, 3, 0, 1)))
    input_data.append(np.transpose(input1_data, (2, 3, 0, 1)))
    input_data.append(np.transpose(input2_data, (2, 3, 0, 1)))
    boxes, classes, scores = yolov5_post_process(input_data)
    img_1 = cv2.cvtColor(img, cv2.COLOR_RGB2BGR)
    if boxes is not None:   
       draw(img_1, boxes, scores, classes) 
       print('-----------')
       cv2.imshow("camera", img_1)
       cv2.waitKey(1)
    #cv2.imshow("listener", img)
    else:
        cv2.imshow("camera", img_1)
        cv2.waitKey(1)
    rknn_lite.release()  
def listener():
    rospy.Subscriber("/rtsp_camera_relay/image", Image, callback)
    rospy.spin()

if __name__ == '__main__':
    listener()
识别结果：