1）新增M027:建筑物下行人检测及计数 2）人群计数及车牌代码优化

AI.py

@@ -735,30 +735,87 @@ def AI_process_Ocr(im0s,modelList,device,detpar):
 
 def AI_process_Crowd(im0s,model,device,postPar):
     timeMixPost = ':0 ms'
-    new_device = torch.device(device)
     time0 = time.time()
-    preds = model.eval(im0s[0])
+    h, w = im0s[0].shape[:2]
+    Wrate = w // 128 * 128 / w
+    Hrate = h // 128 * 128 / h
+    detM = np.array([])
+    if len(model) == 1:
+        detP = model[0].eval(im0s[0])
+        p_postPar = postPar[0]
+
+    else:
+        detP = model[1].eval(im0s[0])
+        p_postPar = postPar[1]
+        detM = np.array(model[0].eval(im0s[0])[0])
+        # d_postPar = postPar[0]
     time1 = time.time()
-    outputs_scores = torch.nn.functional.softmax(preds['pred_logits'], -1)[:, :, 1][0]
+    outputs_scores = torch.nn.functional.softmax(detP['pred_logits'], -1)[:, :, 1][0]
 
-    outputs_points = preds['pred_points'][0]
+    outputs_points = detP['pred_points'][0]
+
+    detP = outputs_points[outputs_scores > p_postPar['conf']].detach().cpu().numpy()
+    detP[:, 0] = detP[:, 0] / Wrate
+    detP[:, 1] = detP[:, 1] / Hrate
+
+    dets = []
+    if not detM.size==0:
+        P = detP.shape[0]  # 第二类边界框数量
+        if  P == 0:
+            time_info = 'No Valid object find'
+            return [im0s[0], im0s[0], dets, 0], time_info
+        x1 = detP[:, 0]  - 5
+        y1 = detP[:, 1]  - 5
+        x2 = detP[:, 0]  + 5
+        y2 = detP[:, 1]  + 5
+        detP = np.column_stack((x1, y1, x2, y2))
+        detM = detM[:,:4]
+        a_x1, a_y1, a_x2, a_y2 = detM[:, 0], detM[:, 1], detM[:, 2], detM[:, 3]
+        b_x1, b_y1, b_x2, b_y2 = detP[:, 0], detP[:, 1], detP[:, 2], detP[:, 3]
+
+        # 扩展维度以便广播 [N,1]
+        a_x1 = a_x1[:, np.newaxis]
+        a_y1 = a_y1[:, np.newaxis]
+        a_x2 = a_x2[:, np.newaxis]
+        a_y2 = a_y2[:, np.newaxis]
+
+        # 计算相交区域坐标 [N,M]
+        inter_x1 = np.maximum(a_x1, b_x1)
+        inter_y1 = np.maximum(a_y1, b_y1)
+        inter_x2 = np.minimum(a_x2, b_x2)
+        inter_y2 = np.minimum(a_y2, b_y2)
+        # 计算相交区域面积 [N,M]
+        inter_area = np.maximum(0, inter_x2 - inter_x1) * np.maximum(0, inter_y2 - inter_y1)
+
+        # 计算每个A框与多少个B框相交（相交面积>0）
+        counts = np.sum(inter_area > 0, axis=1).astype(int)
+
+        # 构建结果数组：前4列是A框坐标，第5列是相交计数
+        detM = np.hstack((detM, counts.reshape(-1, 1)))
+        detM = detM[detM[:, -1] != 0]
+        # 找出所有与A框相交的B框索引
+        p_preds_index = np.any(inter_area > 0, axis=0)
+        p_indices = np.where(p_preds_index)[0]
+        # 提取与第一类相交的第二类边界框
+        if len(p_indices) > 0:
+            detP = detP[p_indices]
+            detP = detP[:, :2]
+            detP[:, 0] = detP[:, 0] + 5
+            detP[:, 1] = detP[:, 1] + 5
+
+            dets= [[detM.tolist(), detP.tolist()]]
+    else:
+        dets = [[detP.tolist()]]
+
+    # for b in detM:
+    #     b_label = '该建筑物下行人及数量：%d'%(int(b[4]))
 
-    points = outputs_points[outputs_scores > postPar['conf']].detach().cpu().numpy().tolist()
-    predict_cnt = int((outputs_scores > postPar['conf']).sum())
-    #img_to_draw = cv2.cvtColor(np.array(img_raw), cv2.COLOR_RGB2BGR)
     time2 = time.time()
-    # for p in points:
-    #     img_to_draw = cv2.circle(img_to_draw, (int(p[0]), int(p[1])), line, (0, 0, 255), -1)
-    Calc_label = '当前人数: %d' % (predict_cnt)
 
-
-    dets = [[Calc_label, points]]
     time_info = 'det:%.1f , post:%.1f  ,timeMixPost:%s ' % (
-    (time1 - time0) * 1000, (time2 - time1) * 1000, timeMixPost)
+        (time1 - time0) * 1000, (time2 - time1) * 1000, timeMixPost)
 
-
-
-    return [im0s[0],im0s[0],dets,0],time_info
+    return [im0s[0], im0s[0], dets, 0], time_info
 
 
 def main():