def test_SquareCompactness(self):
self.df_buildings["sqcom"] = mm.SquareCompactness(
self.df_buildings).series
check = ((4 * math.sqrt(self.df_buildings.geometry.area[0])) /
(self.df_buildings.geometry.length[0]))**2
assert self.df_buildings["sqcom"][0] == check
self.df_buildings["sqcom2"] = mm.SquareCompactness(
self.df_buildings,
areas=self.df_buildings.geometry.area,
perimeters=self.df_buildings.geometry.length,
).series
assert self.df_buildings["sqcom2"][0] == check
def time_SquareCompactness(self):
mm.SquareCompactness(self.df_buildings)
def retrieval_similar_cases(data_path, cases_path, threshold):
threshold = json.loads(threshold)['threshold']
data = geopandas.read_file(data_path) # print(data.shape) #(1, 46)
cases = geopandas.read_file(cases_path) # print(len(cases)) #1055
sf_cases = shapefile.Reader(cases_path, encoding='gbk')
sf_data = shapefile.Reader(data_path, encoding='gbk')
change_list = ['bank', 'hospital', 'mall', 'school', 'subway']
for l in change_list:
cases.loc[cases[l].notnull(), l] = 1
cases.loc[cases[l].isnull(), l] = 0
cases = cases.fillna(value='nan')
data['area'] = momepy.Area(data).series
data['length'] = momepy.Perimeter(data).series
data['ccd_means'] = momepy.CentroidCorners(data).mean
data['ccd_std_stdev'] = momepy.CentroidCorners(data).std
data['circ_comp'] = momepy.CircularCompactness(data).series # 周长紧凑度
data['cwa'] = momepy.CompactnessWeightedAxis(data).series # 紧凑度加权轴
data['convexity'] = momepy.Convexity(data).series # 凸度
data['corners'] = momepy.Corners(data).series # 角数
data['elongation'] = momepy.Elongation(data).series # 伸长率
data['eri'] = momepy.EquivalentRectangularIndex(data).series # 等校矩形指数
data['fractal'] = momepy.FractalDimension(data).series # 分形维数
data['rectangularity'] = momepy.Rectangularity(data).series # 矩形度
data['squ_comp'] = momepy.SquareCompactness(data).series # 紧凑度指数
data['long_ax'] = momepy.LongestAxisLength(data).series # 最长轴的长度值
data['shape_index'] = momepy.ShapeIndex(
data, longest_axis='long_ax').series # 形状索引
cases['area'] = momepy.Area(cases).series
cases['length'] = momepy.Perimeter(cases).series
cases['ccd_means'] = momepy.CentroidCorners(cases).mean
cases['ccd_std_stdev'] = momepy.CentroidCorners(cases).std
cases['circ_comp'] = momepy.CircularCompactness(cases).series # 周长紧凑度
cases['cwa'] = momepy.CompactnessWeightedAxis(cases).series # 紧凑度加权轴
cases['convexity'] = momepy.Convexity(cases).series # 凸度
cases['corners'] = momepy.Corners(cases).series # 角数
cases['elongation'] = momepy.Elongation(cases).series # 伸长率
cases['eri'] = momepy.EquivalentRectangularIndex(cases).series # 等校矩形指数
cases['fractal'] = momepy.FractalDimension(cases).series # 分形维数
cases['rectangularity'] = momepy.Rectangularity(cases).series # 矩形度
cases['squ_comp'] = momepy.SquareCompactness(cases).series # 紧凑度指数
cases['long_ax'] = momepy.LongestAxisLength(cases).series # 最长轴的长度值
cases['shape_index'] = momepy.ShapeIndex(
cases, longest_axis='long_ax').series # 形状索引
test_x = cases.iloc[:, 18:].sub(data.iloc[0, 2:],
axis=1).abs().astype('float')
ori = []
dft = []
him = []
for i in range(len(cases)):
# Ori
ori.append(
abs(
compute_ori(sf_cases.shape(i)) -
compute_ori(sf_data.shape(0))))
# DFT
fd_c, final_x_c, final_y_c = compute_DFT(sf_cases.shape(i))
fd_d, final_x_d, final_y_d = compute_DFT(sf_data.shape(0))
tmp = 0
for k in range(20):
tmp += math.pow((fd_c[k] - fd_d[k]), 2)
dft.append(math.sqrt(tmp))
# him
inter = np.concatenate(
(final_x_c[:, np.newaxis], final_y_c[:, np.newaxis]),
1) # nom_x[:, np.newaxis]新增一个维度
inter = inter.reshape(
len(final_x_c), 1,
2) # !!!OpenCV找轮廓后，返回的ndarray的维数是(100, 1, 2)！！！而不是我们认为的(100, 2)。
inter_d = np.concatenate(
(final_x_d[:, np.newaxis], final_y_d[:, np.newaxis]),
1) # nom_x[:, np.newaxis]新增一个维度
inter_d = inter_d.reshape(
len(final_x_d), 1,
2) # !!!OpenCV找轮廓后，返回的ndarray的维数是(100, 1, 2)！！！而不是我们认为的(100, 2)。
him.append(cv2.matchShapes(inter, inter_d, 1, 0))
test_x['area'] = test_x['area'] * 0.000001
test_x['length'] = test_x['length'] * 0.001
test_x['ori'] = ori
test_x['dft'] = dft
test_x['shape'] = him
loaded_model = pickle.load(open('xgb.pickle.dat', 'rb'))
xgb_pred = loaded_model.predict_proba(test_x)
shutil.rmtree('result')
os.mkdir('result')
result = {}
for i in range(len(xgb_pred)):
if test_x['area'][i] < data['area'][0] * 0.25 and test_x['ori'][
i] < 30 and xgb_pred[i][1] > threshold: # 0.99,26
information = {
'编号ID': float(cases['ID'][i]),
'地块名字': cases['NAME'][i],
'地块所在地': cases['city'][i],
'地块面积': cases['area'][i],
'地块朝向': compute_ori(sf_cases.shape(i)),
'容积率': cases['plot_area'][i],
'价格': cases['price'][i],
'绿化率': cases['greening_r'][i],
'建成日期': cases['build_date'][i],
'school': float(cases['school'][i]),
'mall': float(cases['mall'][i]),
'restaurant': float(cases['restaurant'][i]),
'hospital': float(cases['hospital'][i]),
'subway': float(cases['subway'][i]),
'bank': float(cases['bank'][i]),
'park': float(cases['includ_g'][i]),
'water': float(cases['includ_w'][i])
}
result[i] = json.dumps(information,
sort_keys=True,
indent=4,
separators=(',', ': '),
ensure_ascii=False)
a = 1024
b = 512
img = np.zeros((a, a, 3))
img.fill(255)
landuse = sf_cases
l_shape = landuse.shape(i)
l_convex = Polygon(l_shape.points).convex_hull
x_c = l_convex.centroid.xy[0][
0] # l_convex.centroid.xy:(array('d', [12945692.760656377]), array('d', [4861576.219346005]))
y_c = l_convex.centroid.xy[1][0]
l_dot = np.array(l_shape.points)
l_nom_x = np.array(list(map(int, l_dot[:, 0]))) - int(x_c)
l_nom_y = np.array(list(map(int, l_dot[:, 1]))) - int(y_c)
l_inter = np.concatenate((l_nom_x[:, np.newaxis] + b,
minus(l_nom_y)[:, np.newaxis] + b),
1) # nom_x[:, np.newaxis]新增一个维度
cv2.polylines(
img, [np.asarray(l_inter)], True, (0, 0, 255),
1) # cv2.polylines(画布，点坐标列表，封闭，颜色，宽度)polylines点坐标不能出现浮点型，需要是整型
cv2.imwrite('./result/' + str(cases['ID'][i]) + '.jpg', img)
final_data = json.dumps(result,
sort_keys=True,
indent=4,
separators=(',', ': '),
ensure_ascii=False)
print(final_data)
sf_cases.close()
sf_data.close()
del cases
del data
return final_data