識(shí)別框架還是沿用之前的了MediaPipe Hand。

背景知識(shí)不摘重復(fù)，參見(jiàn)之前的：亞博microros小車-原生ubuntu支持系列：10-畫(huà)筆-CSDN博客

手指控制

src/yahboom_esp32_mediapipe/yahboom_esp32_mediapipe/目錄下新建文件10_HandCtrl.py，代碼如下：

#!/usr/bin/env python3
# encoding: utf-8
import math
import time
import cv2 as cv
import numpy as np
import mediapipe as mp
import rclpy
from rclpy.node import Node
from cv_bridge import CvBridge
from sensor_msgs.msg import Image, CompressedImagefrom rclpy.time import Time
import datetimevolPer = value = index = 0
effect = ["color", "thresh", "blur", "hue", "enhance"]
volBar = 400
class handDetector:def __init__(self, mode=False, maxHands=2, detectorCon=0.5, trackCon=0.5):self.tipIds = [4, 8, 12, 16, 20]self.mpHand = mp.solutions.handsself.mpDraw = mp.solutions.drawing_utilsself.hands = self.mpHand.Hands(#模型初始化static_image_mode=mode,max_num_hands=maxHands,min_detection_confidence=detectorCon,min_tracking_confidence=trackCon)self.lmDrawSpec = mp.solutions.drawing_utils.DrawingSpec(color=(0, 0, 255), thickness=-1, circle_radius=15)self.drawSpec = mp.solutions.drawing_utils.DrawingSpec(color=(0, 255, 0), thickness=10, circle_radius=10)#距離計(jì)算def get_dist(self, point1, point2):x1, y1 = point1x2, y2 = point2return abs(math.sqrt(math.pow(abs(y1 - y2), 2) + math.pow(abs(x1 - x2), 2)))#計(jì)算角度def calc_angle(self, pt1, pt2, pt3):point1 = self.lmList[pt1][1], self.lmList[pt1][2]point2 = self.lmList[pt2][1], self.lmList[pt2][2]point3 = self.lmList[pt3][1], self.lmList[pt3][2]a = self.get_dist(point1, point2)b = self.get_dist(point2, point3)c = self.get_dist(point1, point3)try:#余弦定理radian = math.acos((math.pow(a, 2) + math.pow(b, 2) - math.pow(c, 2)) / (2 * a * b))angle = radian / math.pi * 180#弧度轉(zhuǎn)角度except:angle = 0return abs(angle)def findHands(self, frame, draw=True):img = np.zeros(frame.shape, np.uint8)img_RGB = cv.cvtColor(frame, cv.COLOR_BGR2RGB)#圖像格式轉(zhuǎn)換self.results = self.hands.process(img_RGB)#檢測(cè)if self.results.multi_hand_landmarks:for handLms in self.results.multi_hand_landmarks:#輸出關(guān)鍵點(diǎn)if draw: self.mpDraw.draw_landmarks(img, handLms, self.mpHand.HAND_CONNECTIONS)return imgdef findPosition(self, frame, draw=True):self.lmList = []if self.results.multi_hand_landmarks:for id, lm in enumerate(self.results.multi_hand_landmarks[0].landmark):# print(id,lm)h, w, c = frame.shapecx, cy = int(lm.x * w), int(lm.y * h)# print(id, lm.x, lm.y, lm.z)self.lmList.append([id, cx, cy])#記錄指點(diǎn)序號(hào)與坐標(biāo)if draw: cv.circle(frame, (cx, cy), 15, (0, 0, 255), cv.FILLED)return self.lmListdef frame_combine(slef,frame, src):if len(frame.shape) == 3:frameH, frameW = frame.shape[:2]srcH, srcW = src.shape[:2]dst = np.zeros((max(frameH, srcH), frameW + srcW, 3), np.uint8)dst[:, :frameW] = frame[:, :]dst[:, frameW:] = src[:, :]else:src = cv.cvtColor(src, cv.COLOR_BGR2GRAY)frameH, frameW = frame.shape[:2]imgH, imgW = src.shape[:2]dst = np.zeros((frameH, frameW + imgW), np.uint8)dst[:, :frameW] = frame[:, :]dst[:, frameW:] = src[:, :]return dstclass MY_Picture(Node):def __init__(self, name):super().__init__(name)self.bridge = CvBridge()self.sub_img = self.create_subscription(CompressedImage, '/espRos/esp32camera', self.handleTopic, 1) #獲取esp32傳來(lái)的圖像self.hand_detector = handDetector()self.volPer = self.value = self.index = 0self.effect = ["color", "thresh", "blur", "hue", "enhance"]self.volBar = 400self.last_stamp = Noneself.new_seconds = 0self.fps_seconds = 1def handleTopic(self, msg):self.last_stamp = msg.header.stamp  if self.last_stamp:total_secs = Time(nanoseconds=self.last_stamp.nanosec, seconds=self.last_stamp.sec).nanosecondsdelta = datetime.timedelta(seconds=total_secs * 1e-9)seconds = delta.total_seconds()*100if self.new_seconds != 0:self.fps_seconds = seconds - self.new_secondsself.new_seconds = seconds#保留這次的值start = time.time()frame = self.bridge.compressed_imgmsg_to_cv2(msg)frame = cv.resize(frame, (640, 480))action = cv.waitKey(1) & 0xFFimg  = self.hand_detector.findHands(frame)lmList = self.hand_detector.findPosition(frame, draw=False)if len(lmList) != 0:angle = self.hand_detector.calc_angle(4, 0, 8)#計(jì)算拇指，0點(diǎn)，食指尖的角度x1, y1 = lmList[4][1], lmList[4][2]x2, y2 = lmList[8][1], lmList[8][2]cx, cy = (x1 + x2) // 2, (y1 + y2) // 2cv.circle(img, (x1, y1), 15, (255, 0, 255), cv.FILLED)cv.circle(img, (x2, y2), 15, (255, 0, 255), cv.FILLED)cv.line(img, (x1, y1), (x2, y2), (255, 0, 255), 3)cv.circle(img, (cx, cy), 15, (255, 0, 255), cv.FILLED)if angle <= 10: cv.circle(img, (cx, cy), 15, (0, 255, 0), cv.FILLED)self.volBar = np.interp(angle, [0, 70], [400, 150])self.volPer = np.interp(angle, [0, 70], [0, 100])self.value = np.interp(angle, [0, 70], [0, 255])# print("angle: {},self.value: {}".format(angle, self.value))print(f'mode：{self.effect[self.index]}')# 進(jìn)行閾值二值化操作，大于閾值value的，使用255表示，小于閾值value的，使用0表示if self.effect[self.index]=="thresh":gray = cv.cvtColor(frame, cv.COLOR_BGR2GRAY)frame = cv.threshold(gray, self.value, 255, cv.THRESH_BINARY)[1]# 進(jìn)行高斯濾波,(21, 21)表示高斯矩陣的長(zhǎng)與寬都是21，標(biāo)準(zhǔn)差取valueelif self.effect[self.index]=="blur":frame = cv.GaussianBlur(frame, (21, 21), np.interp(self.value, [0, 255], [0, 11]))# 色彩空間的轉(zhuǎn)化,HSV轉(zhuǎn)換為BGRelif self.effect[self.index]=="hue":frame = cv.cvtColor(frame, cv.COLOR_BGR2HSV)frame[:, :, 0] += int(self.value)frame = cv.cvtColor(frame, cv.COLOR_HSV2BGR)# 調(diào)節(jié)對(duì)比度elif self.effect[self.index]=="enhance":enh_val = self.value / 40clahe = cv.createCLAHE(clipLimit=enh_val, tileGridSize=(8, 8))lab = cv.cvtColor(frame, cv.COLOR_BGR2LAB)lab[:, :, 0] = clahe.apply(lab[:, :, 0])frame = cv.cvtColor(lab, cv.COLOR_LAB2BGR)if action == ord('f'):self.index += 1if self.index >= len(self.effect): self.index = 0end = time.time()fps = 1 / ((end - start)+self.fps_seconds)text = "FPS : " + str(int(fps))cv.putText(frame, text, (20, 30), cv.FONT_HERSHEY_SIMPLEX, 0.9, (0, 0, 255), 1)cv.rectangle(img, (50, 150), (85, 400), (255, 0, 0), 3)cv.rectangle(img, (50, int(self.volBar)), (85, 400), (0, 255, 0), cv.FILLED)cv.putText(img, f'{int(self.volPer)}%', (40, 450), cv.FONT_HERSHEY_COMPLEX, 1, (0, 255, 0), 3)dst = self.hand_detector.frame_combine(frame, img)cv.imshow('dst', dst)def main():print("start it")rclpy.init()esp_img = MY_Picture("My_Picture")try:rclpy.spin(esp_img)except KeyboardInterrupt:passfinally:esp_img.destroy_node()rclpy.shutdown()

訂閱esp32傳出來(lái)的圖像后，通過(guò)MediaPipe去做相關(guān)的識(shí)別后，再通過(guò)記錄手指的點(diǎn)坐標(biāo)，計(jì)算角4-0-8之間度數(shù)。本節(jié)與之前不同，增加了opencv輸出的格式，"color", "thresh", "blur", "hue", "enhance"。默認(rèn)是color，還有閾值化輸出，高斯模糊等其他效果。按F鍵切換

構(gòu)建后運(yùn)行：

ros2 run yahboom_esp32_mediapipe HandCtrl

效果如下：

手勢(shì)識(shí)別

src/yahboom_esp32_mediapipe/yahboom_esp32_mediapipe/目錄下新建文件11_GestureRecognition.py，代碼如下

#!/usr/bin/env python3
# encoding: utf-8
import math
import time
import cv2 as cv
import numpy as np
import mediapipe as mp
import rclpy
from rclpy.node import Node
from cv_bridge import CvBridge
from sensor_msgs.msg import Image, CompressedImagefrom rclpy.time import Time
import datetimeclass handDetector:def __init__(self, mode=False, maxHands=2, detectorCon=0.5, trackCon=0.5):self.tipIds = [4, 8, 12, 16, 20]self.mpHand = mp.solutions.handsself.mpDraw = mp.solutions.drawing_utilsself.hands = self.mpHand.Hands(static_image_mode=mode,max_num_hands=maxHands,min_detection_confidence=detectorCon,min_tracking_confidence=trackCon)self.lmList = []self.lmDrawSpec = mp.solutions.drawing_utils.DrawingSpec(color=(0, 0, 255), thickness=-1, circle_radius=6)self.drawSpec = mp.solutions.drawing_utils.DrawingSpec(color=(0, 255, 0), thickness=2, circle_radius=2)def get_dist(self, point1, point2):x1, y1 = point1x2, y2 = point2return abs(math.sqrt(math.pow(abs(y1 - y2), 2) + math.pow(abs(x1 - x2), 2)))def calc_angle(self, pt1, pt2, pt3):point1 = self.lmList[pt1][1], self.lmList[pt1][2]point2 = self.lmList[pt2][1], self.lmList[pt2][2]point3 = self.lmList[pt3][1], self.lmList[pt3][2]a = self.get_dist(point1, point2)b = self.get_dist(point2, point3)c = self.get_dist(point1, point3)try:radian = math.acos((math.pow(a, 2) + math.pow(b, 2) - math.pow(c, 2)) / (2 * a * b))angle = radian / math.pi * 180except:angle = 0return abs(angle)def findHands(self, frame, draw=True):self.lmList = []img = np.zeros(frame.shape, np.uint8)img_RGB = cv.cvtColor(frame, cv.COLOR_BGR2RGB)self.results = self.hands.process(img_RGB)if self.results.multi_hand_landmarks:for i in range(len(self.results.multi_hand_landmarks)):if draw: self.mpDraw.draw_landmarks(frame, self.results.multi_hand_landmarks[i], self.mpHand.HAND_CONNECTIONS, self.lmDrawSpec, self.drawSpec)self.mpDraw.draw_landmarks(img, self.results.multi_hand_landmarks[i], self.mpHand.HAND_CONNECTIONS, self.lmDrawSpec, self.drawSpec)for id, lm in enumerate(self.results.multi_hand_landmarks[i].landmark):h, w, c = frame.shapecx, cy = int(lm.x * w), int(lm.y * h)self.lmList.append([id, cx, cy])return frame, imgdef frame_combine(slef,frame, src):if len(frame.shape) == 3:frameH, frameW = frame.shape[:2]srcH, srcW = src.shape[:2]dst = np.zeros((max(frameH, srcH), frameW + srcW, 3), np.uint8)dst[:, :frameW] = frame[:, :]dst[:, frameW:] = src[:, :]else:src = cv.cvtColor(src, cv.COLOR_BGR2GRAY)frameH, frameW = frame.shape[:2]imgH, imgW = src.shape[:2]dst = np.zeros((frameH, frameW + imgW), np.uint8)dst[:, :frameW] = frame[:, :]dst[:, frameW:] = src[:, :]return dstdef fingersUp(self):fingers=[]# Thumbif (self.calc_angle(self.tipIds[0],self.tipIds[0] - 1,self.tipIds[0] - 2) > 150.0) and (self.calc_angle(self.tipIds[0] - 1,self.tipIds[0] - 2,self.tipIds[0] - 3) > 150.0): fingers.append(1)else:fingers.append(0)# 4 fingerfor id in range(1, 5):if self.lmList[self.tipIds[id]][2] < self.lmList[self.tipIds[id] - 2][2]:fingers.append(1)else:fingers.append(0)return fingersdef get_gesture(self):gesture = ""fingers = self.fingersUp()if self.lmList[self.tipIds[0]][2] > self.lmList[self.tipIds[1]][2] and \self.lmList[self.tipIds[0]][2] > self.lmList[self.tipIds[2]][2] and \self.lmList[self.tipIds[0]][2] > self.lmList[self.tipIds[3]][2] and \self.lmList[self.tipIds[0]][2] > self.lmList[self.tipIds[4]][2] : gesture = "Thumb_down"elif self.lmList[self.tipIds[0]][2] < self.lmList[self.tipIds[1]][2] and \self.lmList[self.tipIds[0]][2] < self.lmList[self.tipIds[2]][2] and \self.lmList[self.tipIds[0]][2] < self.lmList[self.tipIds[3]][2] and \self.lmList[self.tipIds[0]][2] < self.lmList[self.tipIds[4]][2] and \self.calc_angle(self.tipIds[1] - 1, self.tipIds[1] - 2, self.tipIds[1] - 3) < 150.0 : gesture = "Thumb_up"if fingers.count(1) == 3 or fingers.count(1) == 4:if fingers[0] == 1 and (self.get_dist(self.lmList[4][1:], self.lmList[8][1:])<self.get_dist(self.lmList[4][1:], self.lmList[5][1:])): gesture = "OK"elif fingers[2] == fingers[3] == 0: gesture = "Rock"elif fingers.count(1) == 3: gesture = "Three"else: gesture = "Four"elif fingers.count(1) == 0: gesture = "Zero"elif fingers.count(1) == 1: gesture = "One"elif fingers.count(1) == 2:if fingers[0] == 1 and fingers[4] == 1: gesture = "Six"elif fingers[0] == 1 and self.calc_angle(4, 5, 8) > 90: gesture = "Eight"elif fingers[0] == fingers[1] == 1 and self.get_dist(self.lmList[4][1:], self.lmList[8][1:]) < 50: gesture = "Heart_single"else: gesture = "Two"elif fingers.count(1)==5:gesture = "Five"if self.get_dist(self.lmList[4][1:], self.lmList[8][1:]) < 60 and \self.get_dist(self.lmList[4][1:], self.lmList[12][1:]) < 60 and \self.get_dist(self.lmList[4][1:], self.lmList[16][1:]) < 60 and \self.get_dist(self.lmList[4][1:], self.lmList[20][1:]) < 60 : gesture = "Seven"if self.lmList[self.tipIds[0]][2] < self.lmList[self.tipIds[1]][2] and \self.lmList[self.tipIds[0]][2] < self.lmList[self.tipIds[2]][2] and \self.lmList[self.tipIds[0]][2] < self.lmList[self.tipIds[3]][2] and \self.lmList[self.tipIds[0]][2] < self.lmList[self.tipIds[4]][2] and \self.calc_angle(self.tipIds[1] - 1, self.tipIds[1] - 2, self.tipIds[1] - 3) > 150.0 : gesture = "Eight"return gestureclass MY_Picture(Node):def __init__(self, name):super().__init__(name)self.bridge = CvBridge()self.sub_img = self.create_subscription(CompressedImage, '/espRos/esp32camera', self.handleTopic, 1) #獲取esp32傳來(lái)的圖像self.hand_detector = handDetector(detectorCon=0.75)self.last_stamp = Noneself.new_seconds = 0self.fps_seconds = 1def handleTopic(self, msg):self.last_stamp = msg.header.stamp  if self.last_stamp:total_secs = Time(nanoseconds=self.last_stamp.nanosec, seconds=self.last_stamp.sec).nanosecondsdelta = datetime.timedelta(seconds=total_secs * 1e-9)seconds = delta.total_seconds()*100if self.new_seconds != 0:self.fps_seconds = seconds - self.new_secondsself.new_seconds = seconds#保留這次的值start = time.time()frame = self.bridge.compressed_imgmsg_to_cv2(msg)frame = cv.resize(frame, (640, 480))cv.waitKey(1) frame, img = self.hand_detector.findHands(frame, draw=False)if len(self.hand_detector.lmList) != 0:totalFingers = self.hand_detector.get_gesture()cv.rectangle(frame, (0, 430), (230, 480), (0, 255, 0), cv.FILLED)cv.putText(frame, str(totalFingers), (10, 470), cv.FONT_HERSHEY_PLAIN, 2, (255, 0, 0), 2)end = time.time()fps = 1 / ((end - start)+self.fps_seconds)text = "FPS : " + str(int(fps))cv.putText(frame, text, (20, 30), cv.FONT_HERSHEY_SIMPLEX, 0.9, (0, 0, 255), 1)dist = self.hand_detector.frame_combine(frame, img)cv.imshow('dist', dist)'''
Zero One Two Three Four Five Six Seven Eight
Ok: OK
Rock: rock
Thumb_up : 點(diǎn)贊
Thumb_down: 拇指向下
Heart_single: 單手比心
'''def main():print("start it")rclpy.init()esp_img = MY_Picture("My_Picture")try:rclpy.spin(esp_img)except KeyboardInterrupt:passfinally:esp_img.destroy_node()rclpy.shutdown()

網(wǎng)上有不少這個(gè)例子，差異點(diǎn)可能在手勢(shì)識(shí)別哪里，前面的hand模型都是一樣的。

根據(jù)你預(yù)計(jì)的指點(diǎn)判斷角度或者漏出的手指組合判斷含義。有的也不太準(zhǔn)確，大部分能識(shí)別。

構(gòu)建后運(yùn)行：ros2 run yahboom_esp32_mediapipe GestureRecognition