Table of Contents
Geographic Information Systems Internship Report 1
- Video processing and data cleaning1
1.1. Processing classified data1
1.2. Processing video data1 - Neural network construction 2
2.1. Data processing 2
2.2. Network construction 2
2.2.1. Vgg network construction 3
2.2.2. Resnet network construction 3
2.3. Final result 3 - Visualization 4
3.1. AntvL7 map visualization 4
3.2. Antv G2 timing diagram display 6
3.2.1. Flow timing diagram 6
3.2.2. Timing diagram with thumbnail axis 6
3.3. Data processing 7 - Collect video and point data APP 7
- Description of workload of team members 8
- Reference 8
1. Video processing and data cleaning
1.1. Processing classified data
Language used: Python.
Read the .csv file corresponding to each emotion, obtain the image file name and corresponding score, and then use the PIL library to read the image according to the file name of the image and convert the image into numpy matrix format. When reading the file, each An emotion file is divided into a training set and a test set at a ratio of 8:2. Finally, the matrix and score of the resulting image are written to the file in binary form by calling the pikle library.
1.2. Process video data.
Language used: Python.
First read the text file in the same directory as the video file to get the video start recording time, then read the CSV file to get the time and point data of each collection, convert the collection time and start time into a timestamp and then subtract them. Get the interval time.
Then use the OpenCv library to intercept each frame corresponding to the video according to the interval. Since the format of the image read by OpenCv is BGR, it needs to be converted to RGB format before saving the image.
Then use the pickle library to write the image and corresponding longitude and latitude to the file.
import csv
import os
import pprint
from PIL import Image
import cv2
import numpy as np
import pickle
#图片文件目录
feelfile='D:\\Users\\GISDATA\\scores\\scores\\wealthy'
#CSV文件，保存了图片名和对应分数
input_file = 'D:\\Users\\GISDATA\\scores\\scores\\auto_save_scores_wealthy.csv'
#输出的pickle文件
output_file_pickle="D:\\Users\\GISDATA\\scores\\scores\\feel_boring"
#将数据随机分为训练集和测试集
#获取对应的图片和分数，将图片转为矩阵
#使用pickle写入到文件中
def write_pictle(input_file,input_dir,output_file,):
read_file = open(input_file, "r")
csv_read=csv.reader(read_file)
result = dict(csv_read) #转化为字典
with open(output_file,'wb') as write_file:
#训练集分数
train_score = []
#训练集图片矩阵
train_img = []
#测试集图片矩阵
test_img=[]
#测试集分数
test_score=[]
length=len(result)
for num in range(int(8*length/10)):
#训练集
(file,score)=result.popitem()
if file=='File':
continue
file_name = input_dir + '\\' + file + '.png' # 得到文件的位置
img = Image.open(file_name) # 读取图片
img = img.convert("RGB")
img= np.array(img)
train_img.append(img)
train_score.append(score)
for num in range(int(2*length/10)):
#测试集
(file, score) = result.popitem()
if file == 'File':
continue
file_name = input_dir + '\\' + file + '.png' # 得到文件的位置
#img = cv2.imread(file_name)
# img=cv2.resize(img, (224, 224), interpolation=cv2.INTER_CUBIC)
# img = img[:, :, :: -1]
img = Image.open(file_name) # 读取图片
img = img.convert("RGB")
img = img.resize((224, 224))
img = np.array(img)
test_img.append(img)
test_score.append(score)
f_img = open(output_file_pickle, 'wb')
pickle.dump((train_img,train_score,test_img,test_score), f_img, 0)
f_img.close()
read_file.close()
#测试文件是否正确
def pickle_load(output_file): # 使用pickle从文件中重构python对象
f = open(output_file, 'rb')
train_imgs,train_scores,test_imgs,test_scores = pickle.load(f)
for img in train_imgs:
#测试：矩阵的维度是否正确
print(img.shape)
img=Image.fromarray(img.astype('uint8')).convert('RGB')
Image._show(img)
img = Image.fromarray(test_imgs[0].astype('uint8')).convert('RGB')
Image._show(img)
#测试数据完整性
#打印测试集和训练集数量
print(len(train_imgs))
print(len(train_scores))
print(len(test_imgs))
print(len(test_scores))
f.close()
if __name__=='__main__':
#处理数据
write_pictle(input_file, feelfile, output_file_pickle)
#测试输出文件
pickle_load(output_file_pickle)
