Classification of Selected Citrus Fruits Based on Color Using Machine Vision System

Figures & data

TABLE 1 Color distances of oranges of different classes

Sample	Class 1	Class 2	Class 3	Result
C1	0.0285	0.0678	0.0655	Class 1
C2	0.0408	0.0512	0.0501	Class 1
C3	0.0372	0.0442	0.0436	Class 1
C4	0.0313	0.0416	0.0399	Class 1
C5	0.0390	0.0504	0.0471	Class 1
C6	0.0466	0.0154	0.0235	Class 2
C7	0.0410	0.0164	0.0199	Class 2
C8	0.0579	0.0190	0.0330	Class 2
C9	0.0587	0.0225	0.0321	Class 2
C10	0.0626	0.0238	0.0300	Class 2
C11	0.0563	0.0204	0.0177	Class 3
C12	0.0550	0.0224	0.0152	Class 3
C13	0.0445	0.0219	0.0121	Class 3
C14	0.0461	0.0270	0.0135	Class 3
C15	0.0479	0.0285	0.0256	Class 3

FIGURE 1 Class references for color distance analysis.

FIGURE 2 Fruit samples for color distance analysis.

FIGURE 3 Class references for LDA and PDF analysis.

FIGURE 4 Fruit samples for LDA and PDF analysis.

TABLE 2 Posterier probabilities of various samples using LDA

Sample	P(Class A)	P(Class B)	P(Class C)	P(Class D)	P(Class E)
L1	0.3749	0.6239	0.0012	0.0000	0.0000
L2	0.9090	0.0910	0.0000	0.0000	0.0000
L3	0.0001	0.0785	0.9214	0.0000	0.0000
L4	0.0000	0.0000	0.0000	0.0000	1.0000
L5	0.0000	0.0000	0.0000	0.0000	1.0000
L6	0.0000	0.0000	0.0000	1.0000	0.0000
L7	0.0000	0.0000	0.0000	1.0000	0.0000
L8	0.7085	0.2914	0.0001	0.0000	0.0000
L9	0.0000	0.0000	0.0000	0.0000	1.0000
L10	0.0575	0.9006	0.0419	0.0000	0.0000
L11	0.6813	0.3186	0.0001	0.0000	0.0000
L12	0.0062	0.6134	0.3804	0.0000	0.0000
L13	0.0000	0.0029	0.9971	0.0000	0.0000
L14	0.0000	0.0000	0.0000	1.0000	0.0000
L15	0.7943	0.2057	0.0000	0.0000	0.0000

Where, P(Class X) is the posterior probability that the sample belongs to Class X.

TABLE 3 Correlation coeffecients of various samples using PDF

Sample	Correlation b/w sample and Class A	Correlation b/w sample and Class B	Correlation b/w sample and Class C	Correlation b/w sample and Class D	Correlation b/w sample and Class E
L1	0.7440	0.8353	0.4185	–0.1257	–0.1368
L2	0.9697	0.2639	0.0389	–0.1312	–0.1445
L3	–0.1013	–0.1010	0.0319	–0.3190	–0.4276
L4	–0.0435	–0.0432	–0.1575	–0.1395	0.1595
L5	–0.0945	–0.0945	–0.2462	0.5794	0.7146
L6	–0.1201	–0.1197	0.3551	0.6280	0.3689
L7	–0.1089	–0.1085	0.2497	0.8045	0.4571
L8	0.9770	0.5357	0.2022	–0.1338	–0.1467
L9	–0.0941	–0.0877	–0.2966	0.2846	0.6997
L10	0.5571	0.9401	0.7087	–0.1661	–0.1770
L11	0.9580	0.5469	0.2117	–0.1290	–0.1414
L12	0.2047	0.8920	0.6606	–0.1196	–0.1275
L13	–0.0213	0.1594	0.5377	–0.1255	–0.1241
L14	–0.1397	–0.1397	–0.0551	0.8995	0.7992
L15	0.9965	0.4147	0.1299	–0.1317	–0.1447

FIGURE 5 Plot of probability distributions of a sample and five classes.

FIGURE 6 Probability distributions of 5 classes—Class A, B, C of oranges, sweet lime, and lemon.

TABLE 4 Results using hue as parameter of classification

Sample number	Manual assessment	Result using LDA	Result using PDF
L1	Class B	Class B	Class B
L2	Class A	Class A	Class A
L3	Class C	Class C	Class C
L4	Class E	Class E	Class E
L5	Class E	Class E	Class E
L6	Class D	Class D	Class D
L7	Class D	Class D	Class D
L8	Class A	Class A	Class A
L9	Class E	Class E	Class E
L10	Class B	Class B	Class B
L11	Class A	Class A	Class A
L12	Class B	Class B	Class B
L13	Class C	Class C	Class C
L14	Class D	Class D	Class D
L15	Class A	Class A	Class A

TABLE 5 Results using saturation as parameter of classification

Sample number	Manual Aassessment	Result using LDA	Result using PDF
L1	Class B	Class C	Class C
L2	Class A	Class C	Class E
L3	Class C	Class C	Class B
L4	Class E	Class E	Class E
L5	Class E	Class E	Class E
L6	Class D	Class D	Class D
L7	Class D	Class D	Class D
L8	Class A	Class C	Class B
L9	Class E	Class E	Class E
L10	Class B	Class C	Class C
L11	Class A	Class C	Class B
L12	Class B	Class C	Class C
L13	Class C	Class C	Class C
L14	Class D	Class D	Class D
L15	Class A	Class C	Class C

FIGURE 7 Sweet-lime samples for maturity analysis.

TABLE 6 Orange fruits classification based on maturity

Sample number	R/G ratio	Hue mean	Hue median	Maturity
C1	1.0707	0.1448	0.0453	Ripe
C2	1.0652	0.1949	0.0467	Ripe
C3	1.0630	0.2149	0.0585	Ripe
C4	1.0548	0.2179	0.0586	Ripe
C5	1.0580	0.2115	0.0523	Ripe
C6	1.0396	0.2294	0.0657	Moderately ripe
C7	1.0401	0.2197	0.0716	Moderately ripe
C8	1.0362	0.2256	0.0812	Moderately ripe
C9	1.0349	0.2350	0.0800	Moderately ripe
C10	1.0384	0.2398	0.0818	Moderately ripe
C11	1.0266	0.2422	0.0931	Unripe
C12	1.0270	0.2467	0.0936	Unripe
C13	1.0293	0.2395	0.0959	Unripe
C14	1.0270	0.2400	0.1031	Unripe
C15	1.0265	0.2433	0.1043	Unripe

TABLE 7 Sweet-lime fruits classification based on maturity

Sample number	R/G ratio	Hue mean	Hue median	Maturity
M1	0.9638	0.1910	0.1830	Ripe
M2	0.9976	0.2145	0.1732	Ripe
M3	0.9837	0.1795	0.1744	Ripe
M4	0.9989	0.1849	0.1688	Ripe
M5	1.0138	0.1731	0.1646	Ripe
M6	0.9484	0.2398	0.1923	Moderately ripe
M7	0.9015	0.2249	0.1993	Moderately ripe
M8	0.8909	0.2237	0.204	Moderately ripe
M9	0.8662	0.2335	0.2197	Moderately ripe
M10	0.8691	0.2286	0.2147	Moderately ripe
M11	0.8463	0.2455	0.2199	Unripe
M12	0.8441	0.2447	0.2208	Unripe
M13	0.7688	0.2501	0.2421	Unripe
M14	0.8355	0.3387	0.2462	Unripe
M15	0.7631	0.3014	0.2485	Unripe

TABLE 8 Criteria for the matutity class

Parameter	Unripe	Moderately ripe	Ripe
Orange fruits
R/G ratio	<1.03	1.03–1.05	>1.05
Hue mean	>0.24	0.22–0.24	<0.22
Hue median	>0.09	0.06–0.09	<0.06
Sweet-lime fruits
R/G ratio	<0.85	0.85–0.95	>0.95
Hue mean	>0.24	0.22–0.24	<0.22
Hue median	>0.22	0.19–0.22	<0.19

FIGURE 8 Algorithms used in the proposed methods of color analysis and maturity analysis.