大佬們好,分享一下我用魯班貓做ros主控,stm32f407做底層驅動的一個ros小車。
目的是識別煙霧并凈化:凈化是用的負離子發生器(效果如文章頂部視頻,凈化還是很頂的),外加扇葉將其擴散出去。同時也具有環境氣體濃度(質量)檢測的功能。
b站鏈接:
https://www.bilibili.com/video/BV1hh4y1n7Fz/?vd_source=4fa660ff7e4423139e6ebdbd4dece6c7
這是我去年12月底開始做的,入坑魯班貓算是比較早了。在讀大三學生。正奧里給考研中。。。
最底下還塞了一塊vet6和一塊esp32.
板子上加了個風扇,為了散熱快。
魯班貓1s做ROS主控用于ros建圖(gmapping)和導航,同時接入NPU做抽煙監測,模型是yolov5自己訓練的模型轉化成rknn部署在板子上。
功能部分即功能層的stm32與串口屏、esp32通信部分。功能層的主要目的是獲取傳感器數據和通過繼電器控制小車前端的負離子發生器和兩個加快負離子擴散的風扇。這里的stm32相當于一個中轉,用的是rt—thread實時操作系統,版本是4.0.2(寫的比較早,當時的rtt還有小bug,現在已經很好用了。)
開啟三個串口:一個用于讀取傳感器,一個用于接收和發送指令給串口屏,一個用于給esp32傳輸數據,通過esp32將數據發送到巴法云平臺,做接入小程序中轉。
篇幅有限,代碼放在了網盤上。
鏈接:https://pan.baidu.com/s/1ltgypPMq9heezk412r4IKw?pwd=jhzs
提取碼:jhzs
因為用的是rtt,移植性很高,故只寫了應用層的main.c函數。如下:
氣體傳感器如下(所用的是串口協議)
#include <rtthread.h>
/ *串口1用來調試* /
#define DBG_TAG \"main\"
#define DBG_LVL DBG_LOG
#include <rtdbg.h>
#include <string.h>
#include <serial.h>//此處有坑,要改頭文件路徑為rt-thread/components/drivers/include/drivers
#include <stdio.h>
#include \"stdlib.h\"
#defineleft_motor_run{rt_pin_write(6,PIN_LOW );rt_pin_write(7,PIN_HIGH);}
#defineleft_motor_back{rt_pin_write(6,PIN_HIGH );rt_pin_write(7,PIN_LOW);}
#definestoping{rt_pin_write(6,PIN_HIGH );rt_pin_write(7,PIN_HIGH);rt_pin_write(16,PIN_HIGH );rt_pin_write(17,PIN_HIGH);}
#defineright_motor_run{rt_pin_write(16,PIN_LOW );rt_pin_write(17,PIN_HIGH);}
#defineright_motor_back{rt_pin_write(16,PIN_HIGH );rt_pin_write(17,PIN_LOW);}
#define key1_openrt_pin_write(51,PIN_LOW );//d3
#define key1_closert_pin_write(51,PIN_HIGH );
#define key2_openrt_pin_write(52,PIN_LOW );//d4
#define key2_closert_pin_write(52,PIN_HIGH );
#define key3_openrt_pin_write(53,PIN_LOW );//d5
#define key3_closert_pin_write(53,PIN_HIGH );
/*micropython esp32與rtt串口DMA傳輸數據時有坑,
* 需在drv_usart.c找到HAL_UART_RxCpltCallback和HAL_UART_RxHalfCpltCallback將dma_isr(&uart->serial)注釋掉,
* 能降低數據錯誤率*/
/ *串口2的變量 115200* /
struct serial_configureuar2_configs = RT_SERIAL_CONFIG_DEFAULT;
rt_sem_t sem2;
rt_device_t uar2_dev;
rt_thread_t uar_2_th;
rt_thread_t uar_2_deal;
char buffer[128] = {0};
rt_size_t rxlen2 = 0;
/ *串口3的變量 9600* /
struct serial_configureuar3_configs = MY_SERIAL_CONFIG_DEFAULT;
rt_sem_t sem3;
rt_device_t uar3_dev;
rt_thread_t uar_3_th;
uint8_t buffer3[17] = {0};
rt_size_t rxlen3 = 0;
/ *串口4的變量 115200* /
struct serial_configureuar4_configs = RT_SERIAL_CONFIG_DEFAULT;
rt_sem_t sem4;
rt_device_t uar4_dev;
rt_thread_t uar_4_th;
rt_uint8_t buffer4[256] = {0xff};
rt_size_t rxlen4 = 0;
//char deal;
rt_uint8_t deal ;
char wheater[8];
char humidity[4];
char temperature[4];
char wind_speed[4];
char shi[3];
char miao[3];
char fen[3];
char wheater_deal[23]=\"main2.g3.txt=\"\";
char humidity_deal[18]=\"main2.g1.txt=\"\";
char temperature_deal[17]=\"main2.g0.txt=\"\";
char wind_speed_deal[19]=\"main2.g2.txt=\"\";
char shi_deal[15] = \"main.z1.val=\";
char miao_deal[15] = \"main.z0.val=\";
char fen_deal[15] = \"main.z2.val=\";
char end[2]=\"\"\";
char xf_end[3];//串口屏控制幀尾
void uar2_thread_entry(void *parameter)//串口2DMA線程入口
{
rt_size_t len = 0;
/*發送ch2o數據*/
rt_device_write(uar4_dev,0,ch2o_date,sizeof(ch2o_date));
rt_device_write(uar4_dev,0 ,xf_end,sizeof(xf_end));
/*end*/
/*發送tvoc數據*/
rt_device_write(uar4_dev,0 ,tvoc_date,sizeof(tvoc_date));
rt_device_write(uar4_dev,0 ,xf_end,sizeof(xf_end));
/*end*/
/*發送pm2.5數據*/
rt_device_write(uar4_dev,0 ,pm2_5_date,sizeof(pm2_5_date));
rt_device_write(uar4_dev,0 ,xf_end,sizeof(xf_end));
/*end*/
/*發送pm10數據*/
rt_device_write(uar4_dev,0 ,pm10_date,sizeof(pm10_date));
rt_device_write(uar4_dev,0 ,xf_end,sizeof(xf_end));
/*end*/
/*發送temp數據*/rt_device_write(uar4_dev,0,temp_date,sizeof(temp_date));
rt_device_write(uar4_dev,0 ,xf_end,sizeof(xf_end));
/*end*/
/*發送humi數據*/
rt_device_write(uar4_dev,0,humi_date,sizeof(humi_date));
rt_device_write(uar4_dev,0 ,xf_end,sizeof(xf_end));
/*end*/
memset(buffer4, 0, sizeof buffer4);
// }
}
}
rt_err_t uar3_rxback(rt_device_t dev, rt_size_t size)//串口3接收//回調函數9600
{
rxlen3 = size;
rt_sem_release(sem3);
return RT_EOK;
}
void uar4_thread_entry(void *parameter)//串口4DMA線程入口
{
rt_size_t len = 0;
while(1)
{
rt_sem_take(sem4, RT_WAITING_FOREVER);
len = rt_device_read(uar4_dev, 0, buffer4, rxlen4);
buffer4[len] = \'\\\\0\';
rt_kprintf(\"%c\\\\n\",len);
if (buffer4[0] == 0x02) {
deal=0x02;
rt_device_write(uar2_dev,0 ,&deal,16);
rt_kprintf(\"uart4\");
deal=0;
}
if (buffer4[0] == 0x01) {
deal=0x01;
rt_device_write(uar2_dev,0 ,&deal,16);
rt_kprintf(\"uart4\");
deal=0;
}
if (buffer4[0] == 0x03) {
deal=0x03;
//memset(buffer,0,sizeof(buffer));
rt_device_write(uar2_dev,0 ,&deal,16);
rt_kprintf(\"uart4\");
deal=0;
}
switch(buffer4[0])
{
case 0x44: key1_open;break;
case 0x55: key1_close;break;
case 0x66: key2_open;break;
case 0x77: key2_close;break;
case 0x88: key3_open;break;
case 0x99: key3_close;break;
}
}
}
rt_err_t uar4_rxback(rt_device_t dev, rt_size_t size)//串口4接收回調函數
{
rxlen4 = size;
rt_sem_release(sem4);
return RT_EOK;
}
void clearmachine_and_motor_pin_init()
{
rt_pin_mode(6,PIN_MODE_OUTPUT );//a6
rt_pin_mode(7,PIN_MODE_OUTPUT );//a7
rt_pin_mode(16,PIN_MODE_OUTPUT );//b0
rt_pin_mode(17,PIN_MODE_OUTPUT );//b1
rt_pin_mode(28, PIN_MODE_INPUT);//b12 left
rt_pin_mode(29,PIN_MODE_INPUT);//b13right
rt_pin_mode(51,PIN_MODE_OUTPUT );//d3
rt_pin_mode(52,PIN_MODE_OUTPUT );//d4
rt_pin_mode(53,PIN_MODE_OUTPUT );//d5
key1_close;
key2_close;
key3_close;
}
int main(void)
{
clearmachine_and_motor_pin_init();//引腳初始化
/*串口2 DMA初始化*/
uar2_dev = rt_device_find(\"uart2\");
if (uar2_dev == NULL) {
LOG_E(\"rt_device_find[uart2] FAILED...\\\\\\\\\\\\\\\\n\");
return -EINVAL;
}
rt_device_open(uar2_dev, RT_DEVICE_OFLAG_RDWR | RT_DEVICE_FLAG_DMA_RX);
rt_device_control(uar2_dev, RT_DEVICE_CTRL_CONFIG, (void *)&uar2_configs);
rt_device_set_rx_indicate(uar2_dev, uar2_rxback);
uar_2_th = rt_thread_create(\"uar2_rx_thread\", uar2_thread_entry, NULL, 4096, 10, 5);
rt_thread_startup(uar_2_th);
sem2 = rt_sem_create(\"sem2\", 1, RT_IPC_FLAG_FIFO);
if(sem2 == RT_NULL){
LOG_E(\"sem2 rt_sem_create failed...\\\\\\\\\\\\\\\\n\");
return -ENOMEM;
}
LOG_D(\"sem2 rt_sem_create successed...\\\\\\\\\\\\\\\\n\");
/*串口2 DMA初始化結束*/
//uar_2_deal = rt_thread_create(\"uar2_deal\", uar2_deal_entry, NULL, 512, 13, 5);
//rt_thread_startup(uar_2_deal);
/*串口3 DMA初始化9600*/
uar3_dev = rt_device_find(\"uart3\");
if (uar3_dev == NULL) {
LOG_E(\"rt_device_find[uart3] FAILED...\\\\\\\\\\\\\\\\n\");
return -EINVAL;
}
rt_device_open(uar3_dev, RT_DEVICE_OFLAG_RDWR | RT_DEVICE_FLAG_DMA_RX);
rt_device_control(uar3_dev, RT_DEVICE_CTRL_CONFIG, (void *)&uar3_configs);
rt_device_set_rx_indicate(uar3_dev, uar3_rxback);
uar_3_th = rt_thread_create(\"uar3_rx_thread\", uar3_thread_entry, NULL, 4096, 12, 5);
rt_thread_startup(uar_3_th);
sem3 = rt_sem_create(\"sem3\", 1, RT_IPC_FLAG_FIFO);
if(sem3 == RT_NULL){
LOG_E(\"sem3 rt_sem_create failed...\\\\\\\\\\\\\\\\n\");
return -ENOMEM;
}
LOG_D(\"sem3 rt_sem_create successed...\\\\\\\\\\\\\\\\n\");
/*串口3 DMA初始化結束*/
/*串口4 DMA初始化*/
uar4_dev = rt_device_find(\"uart4\");
if (uar4_dev == NULL) {
LOG_E(\"rt_device_find[uart4] FAILED...\\\\\\\\\\\\\\\\n\");
return -EINVAL;
}
rt_device_open(uar4_dev, RT_DEVICE_OFLAG_RDWR | RT_DEVICE_FLAG_DMA_RX);
rt_device_control(uar4_dev, RT_DEVICE_CTRL_CONFIG, (void *)&uar4_configs);
rt_device_set_rx_indicate(uar4_dev, uar4_rxback);
uar_4_th = rt_thread_create(\"uar4_rx_thread\", uar4_thread_entry, NULL, 4096, 11, 5);
rt_thread_startup(uar_4_th);
sem4 = rt_sem_create(\"sem4\", 4, RT_IPC_FLAG_FIFO);
if(sem4 == RT_NULL){
LOG_E(\"sem4 rt_sem_create failed...\\\\\\\\\\\\\\\\n\");
return -ENOMEM;
}
LOG_D(\"sem4 rt_sem_create successed...\\\\\\\\\\\\\\\\n\");
/*串口4 DMA初始化結束*/
return RT_EOK;
}
最后是串口屏顯示,同時數據也能在微信小程序上查看
接下來是魯班貓1s做ros主控的部分。
1、移植輪趣大佬的ros源碼:
根據我現有的硬件:思嵐a1雷達、一個usb rgb攝像頭選擇合適的功能包,然后開始移植。
中途會出現很多錯誤。例如缺少部分功能包,sudo apt install ros-noetic-(包名)【我的ros版本是noetic】。
2、移植完畢后發現大佬們并沒有使用魯班貓上的npu。所以我嘗試了用npu跑yolov5在debain10的環境下用python接口效果如下:
Python與c++接口將圖片監測改成實時攝像頭的代碼:
只需更改cv.Capture()函數的攝像頭設備號即可。
鏈接:https://pan.baidu.com/s/1gauOezF-X8ZuvU4b0I4v4A?pwd=jhzs
提取碼:jhzs
Python接口的yolov7只需更改yolov5代碼的錨點即可
以下只列出主函數部分,完整的在鏈接里。
import urllib
import time
import sys
import numpy as np
import cv2
from rknnlite.api import RKNNLite
#from PIL import Image
RKNN_MODEL = \'mask.rknn\'
IMG_PATH = \'./test.jpg\'
OBJ_THRESH = 0.25
NMS_THRESH = 0.45
IMG_SIZE = 640
......(省略中間部分)
if __name__ == \'__main__\':
# Create RKNN object
rknn = RKNNLite()
# init runtime environment
print(\'--> Load RKNN model\')
ret = rknn.load_rknn(RKNN_MODEL)
#ret = rknn.init_runtime(target=\'rv1126\', device_id=\'256fca8144d3b5af\')
if ret != 0:
print(\'Load RKNN model failed\')
exit(ret)
print(\'done\')
ret = rknn.init_runtime()
if ret != 0:
print(\'Init runtime environment failed!\')
exit(ret)
print(\'done\')
capture = cv2.VideoCapture(9)
ref, frame = capture.read()
if not ref:
raise ValueError(\"error reading\")
fps = 0.0
while(True):
t1 = time.time()
#
ref, frame = capture.read()
if not ref:
break
# BGRtoRGB
frame = cv2.cvtColor(frame,cv2.COLOR_BGR2RGB)
#############
img = frame
img, ratio, (dw, dh) = letterbox(img, new_shape=(IMG_SIZE, IMG_SIZE))
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
# Inference
print(\'--> Running model\')
outputs = rknn.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)
#img_1 = img_1[:,:,::-1]
if boxes is not None:
draw(img_1, boxes, scores, classes)
fps= ( fps + (1./(time.time()-t1)) ) / 2
print(\"fps= %.2f\"%(fps))
#img_1 = cv2.putText(frame, \"fps= %.2f\"%(fps), (0, 40), cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 255, 0), 2)
cv2.imshow(\"video\",img_1[:,:,::-1])
c= cv2.waitKey(1) & 0xff
if c==27:
capture.release()
break
print(\"Video Detection Done!\")
capture.release()
cv2.destroyAllWindows()
但這還沒有接到ros中,為此我去翻rknn的github找到了接入ros的方法。
Ros功能包如下:
Launch文件:
Yolov5.launch
<param name=\"model_file\" value=\"yolov5s-640-640.rknn\"/>
<param name=\"display_output\" value=\"$(arg display_output)\"/>
<param name=\"prob_threshold\" value=\"0.35\"/>
<param name=\"chip_type\" value=\"$(arg chip_type)\"/>
<remap from=\"/camera/image_raw\" to=\"$(arg camera_topic)\"/>
Camrea.Launch
<param name=\"video_device\" value=\"/dev/$(arg device)\" />
<param name=\"image_width\" value=\"640\" />
<param name=\"image_height\" value=\"480\" />
<param name=\"framerate\" value=\"30\" />
<param name=\"pixel_format\" value=\"yuyv\" />
<param name=\"camera_frame_id\" value=\"usn_cam\" />
<param name=\"io_method\" value=\"mmap\"/>
<param name=\"camera_name\" value=\"usn_cam\"/>
啟動攝像頭
默認的攝像頭設備號為video0 魯班貓為video9
1、roslaunch rknn_ros camera.launch
2、roslaunch rknn_ros camera.launch device:=video9(可傳參或者改launch)
3、roslaunch rknn_ros yolov5.launch chip_type:=RK3566
鏈接:
鏈接:https://pan.baidu.com/s/1QhfRjDs1sftAB0Q-TS5dBA?pwd=jhzs
提取碼:jhzs
不出意外改好板子型號和對應的video就能用了。
可打開rviz或者rqt_image_view查看。
模型是我自己訓練的,鏈接如下:
鏈接:https://pan.baidu.com/s/1FSJyW6kp4cy3-yakTq_Q4g?pwd=jhzs
提取碼:jhzs
YOLOV5配置和使用:
官方的源碼是不建議的:
用這個:
https://gitcode.net/mirrors/airockchip/yolov5?utm_source=csdn_github_accelerator
這是瑞芯微官方推薦的源碼,但是也需要更改。
yolov5-master\\\\models下的yolo.py
找到
def forward(self, x):
函數,更改為:
def forward(self, x):
z = []# inference output
for i in range(self.nl):
if os.getenv(\'RKNN_model_hack\', \'0\') != \'0\':
z.append(torch.sigmoid(self.m[i](x[i])))
continue
x[i] = self.m[i](x[i])# conv
\'\'\'
bs, _, ny, nx = x[i].shape# x(bs,255,20,20) to x(bs,3,20,20,85)
x[i] = x[i].view(bs, self.na, self.no, ny, nx).permute(0, 1, 3, 4, 2).contiguous()
if not self.training:# inference
if self.onnx_dynamic or self.grid[i].shape[2:4] != x[i].shape[2:4]:
self.grid[i], self.anchor_grid[i] = self._make_grid(nx, ny, i)
y = x[i].sigmoid()
if self.inplace:
y[..., 0:2] = (y[..., 0:2] * 2 + self.grid[i]) * self.stride[i]# xy
y[..., 2:4] = (y[..., 2:4] * 2) ** 2 * self.anchor_grid[i]# wh
else:# for YOLOv5 on AWS Inferentia https://github.com/ultralytics/yolov5/pull/2953
xy, wh, conf = y.split((2, 2, self.nc + 1), 4)# y.tensor_split((2, 4, 5), 4)# torch 1.8.0
xy = (xy * 2 + self.grid[i]) * self.stride[i]# xy
wh = (wh * 2) ** 2 * self.anchor_grid[i]# wh
y = torch.cat((xy, wh, conf), 4)
z.append(y.view(bs, -1, self.no))
if os.getenv(\'RKNN_model_hack\', \'0\') != \'0\':
return z
return x if self.training else (torch.cat(z, 1),) if self.export else (torch.cat(z, 1), x)
\'\'\'
return x[0],x[1],x[2]
這樣就可以在pt權重轉onnx時去掉最后一個Detect層。
pt轉onnx指令
python export.py --weights yolov5s.pt --img 640 --batch 1 --opset 11 --include onnx
紅色字體部分換成要轉換的權重文件例如我的就是:
**python export.py --weights ** **weights/best.pt ** --img 640 --batch 1 --opset 11 --include onnx
opset選擇11。
這樣出來的模型是有三個節點的模型,才是可用的。
可用netron查看:
netron:https://netron.app/(瀏覽器網址)
將模型拖到頁面可查看。
有三個輸出節點。
且要記好三個節點的名字。
在官方要求的ubuntu pc端上進行模型轉換。
我這里有個改好的yolov5源碼(里面是我訓練的抽煙監測模型)
網盤鏈接如下:
鏈接:https://pan.baidu.com/s/1fXKNoXhu4m1SmTr4fc-afg?pwd=jhzs
提取碼:jhzs
Chatgpt部分是b站機器人阿杰github開源項目。
https://www.bilibili.com/video/BV12M4y1R76M/?spm_id_from=333.788
效果如圖:
嗚。。。不要看問得什么
整車的sw模型鏈接:soildwork2020及以上版本可直接打開
鏈接:https://pan.baidu.com/s/1KqB1SOD418dCvyDaZFMgpg?pwd=jhzs
提取碼:jhzs當時還理想化的擼了個履帶,可后來發現打印出來根本用不了,故放棄,換成了輪子。
放鏈接是希望能夠幫到像我一樣步步踩坑的菜鳥級選手。我是老踩坑怪了。
有不當的地方,還望大佬們海涵。
2023-03-24 20:05:14
評論
查看更多