k230 fall detection

K230 and Raspberry Pi communication

1. Experimental Prerequisites

This tutorial uses the Raspberry Pi 5 development board, and the corresponding routine path is [14.export\Raspberrypi-K230\10_k230_falldown_detect.py].

K230 needs to run the [14.export\CanmvIDE-K230\10.falldown_detect.py] program to start the experiment. It is recommended to download it as an offline program.

Things you need:

Windows computer RDK X5 development board USB to TTL module K230 visual module (including TF card with image burned in) Type-C data cable Connection cable

2. Experimental wiring

3. Main code explanation

​x
import serial
​
com="/dev/ttyUSB0"
ser = serial.Serial(com, 115200)
​
FUNC_ID = 10
​
def parse_data(data):
    if data[0] == ord('$') and data[len(data)-1] == ord('#'):
        data_list = data[1:len(data)-1].decode('utf-8').split(',')
        data_len = int(data_list[0])
        data_id = int(data_list[1])
        if data_len == len(data) and data_id == FUNC_ID:
            # print(data_list)
            x = int(data_list[2])
            y = int(data_list[3])
            w = int(data_list[4])
            h = int(data_list[5])
            msg = data_list[6]
            score = int(data_list[7])/100.0
            return x, y, w, h, msg, score
        elif (data_len != len(data)):
            print("data len error:", data_len, len(data))
        elif(data_id != FUNC_ID):
            print("func id error:", data_id, FUNC_ID)
    else:
        print("pto error", data)
    return -1, -1, -1, -1, "", -1
​
while True:
    if ser.in_waiting:
        data = ser.readline()
        # print("rx:", data)
        x, y, w, h, msg, score = parse_data(data.rstrip(b'\n'))
        print("falldown:x:%d, y:%d, w:%d, h:%d, %s, %.2f" % (x, y, w, h, msg, score))

The above program is for parsing K230 data. Only when it complies with specific protocols can the corresponding data be parsed.

in

• x: is the horizontal coordinate of the upper left corner of the recognized box
• y: is the vertical coordinate of the upper left corner of the recognized box
• w: is the width of the recognized frame
• h: is the length of the recognized frame
• state: is the state of recognition. If a fall is detected, it is "Fall". If no fall is detected, it is "NoFall".
• score: is the score of the fall state

4. Experimental Phenomenon

1. After connecting the cables, the k230 visual module runs offline. After K230 is connected to Canmv IDE, open the corresponding program, click [Save open script to CanMV board (as main.py)] on the toolbar, and then restart K230.

2. Transfer the program file to the system, open the terminal and enter the corresponding directory, then run the following command to start the program.

xxxxxxxxxx
python3 10_k230_falldown_detect.py

3. When the K230 camera image recognizes a human body, the terminal will parse and print out the information transmitted by the K230.

in

• x: is the horizontal coordinate of the upper left corner of the recognized box

• y: is the vertical coordinate of the upper left corner of the recognized box

• w: is the width of the recognized frame

• h: is the length of the recognized frame

• state: is the state of recognition. If a fall is detected, it is "Fall". If no fall is detected, it is "NoFall".

• score: is the score of the fall state

  As shown in the figure below